重量级报告!2015中国癌症统计数据

原文:《Cancer statistics in China 2015》,发表于CA-Cancer J Clin,影响因子144.8

  翻译:银河医药团队王明瑞主译13260428811(实习生),张金洋18001315838

随着发病率和死亡率越来越高,癌症正成为中国首要的死亡原因和一个重要的公共卫生问题。因为中国人口多(13.7亿人),先前的国家发病率和死 亡率评估都限制在上世纪90年代的小样本或基于特定的年份。现在通过国家中央癌症登记处的高质量数据,作者分析了72个地区基于人群的癌症登记 (2009-2011),代表了6.5%的人口,用以估计2015年新病例和癌症死亡的人数。趋势分析(2000-2011)使用了22个登记处的数据。 结果表明,2015年预计有4292000个新癌症病例和 2814000个癌症死亡,肺癌的发病率和死亡率都是最高的。胃癌、食道癌和肝癌的发病率和死亡率也很高。将所有癌症的发病率和死亡率结合起来看,农村居 民的年龄标准化数据要高于城镇居民(发病率213.6人/10万人vs191.5人/10万人;死亡率149人/10万人vs109.5人/10万人)。 将所有的癌症结合起来,在2000年到2011年之间,男性的发病率保持稳定(每年+0.2%;P=.1),女性发病率明显上升(每 年+2.2%;P<.05)。

与此相对,死亡率自从 2006 年之后一直在下降,无论男性(每年-1.4%;P<.05)还是女性(每年-1.1%;P<.05)。

很多评估范围内的癌症可以通过减少癌症风险因素、提高临床护理服务的效率,尤其是农村人口和弱势群体,来减少发病率和死亡率。

  1 、介绍

中国的癌症发病率和死亡率一直在上升,从2010年开始已经成为主要的致死原因,成为了中国的一个主要公共卫生问题。这个逐渐增加的压力有相当 大的一部分可以归于人口的增长和老龄化以及社会人口统计的变化。尽管之前有全国发病率的评估,但那些评估或者只能代表很小的人口(小于2%),或者只有特 定的年份。这对评估的不确定性和代表性有影响,并且会潜在的影响癌症控制政策的制定。因为之前的中国癌症预防与控制项目(2004-2010)是10年之 前发布的,所以一个关于中国全国与各地方癌症规模与概况的更加复杂的描绘将会提供更加清晰的优先顺序,为制定基于癌症谱的政策和规划制定提供参考,减轻国 家的癌症负担。

这篇研究评估了全国范围内的癌症发病率、死亡率和存活率;几个主要癌症分区域的发病率和死亡率;几个主要癌症的时间趋势以及这个信息对中国癌症防控提供的指导。

  2 、数据源和方法

2.1 中国癌症登记

全国肿瘤登记中心(NCCR)建立于2002年,负责收集、评估、发布中国的癌症数据。癌症诊断会被上报给地方癌症登记,这些数据有多个来源, 包括地方医院和社区卫生中心,以及城镇居民基本医疗保险和新农合。自2002年起,标准登记条例的实施已经大大提升了癌症数据的质量。2008年,卫生部 通过中央财政体系实施了国家癌症登记项目。自此之后,各地基于人群的登记数量从2008年的54(人口覆盖1.1亿人)上升到了2014年的308(人口 覆盖3亿人)。

目前并非所有的登记都有足够的可供上报的高质量数据。每个地方登记处提交的数据都会受到 NCCR 和国际癌症研究机构/国际癌症注册协会(IARC/IACR)的检查。质量评估包括但不限于,形态学鉴定所占比例(MV%),有死亡证明的癌症病例所占比 例(DCO%),死亡率/发病率比率(M/I),未认证癌症的比例(UB%),不明确或位置原发癌部位的百分比(CPU%)。只有满足以上这些要求的数据 才会在分析中使用。登记数据的具体质量分类可以在之前的出版物中找到(表 1)。数据分类为A或B在这篇研究中被认为是可以接受的。提交的登记数据可供使用的比例随着年份有变动,从2009年的69.2%(104 个中 72 个可用),到2010年的66.2%(219 个中145个可用),和 2011年的75.6%(234个中177个可用)。我们使用了在三年中始终可用的72个登记处的数据。

2.2 癌症发病率数据

为了估计2015年中国新增癌症数量,我们使用72个基于人群的癌症登记点的最近的数据(2009-2011登记的癌症病例)(表1),人口覆 盖8850万人,大概为全国人口的6.5%。表1的更少数量的登记点(22个)提供了4440万人口的覆盖率,有着2000-2011这12年的高质量数 据,用于发病率的时间趋势分析。这两套数据的癌症登记点可以在表1中查找到。

我们这篇研究中不仅包括了侵袭性肿瘤,也包括了参照国际规则定义的多原发癌。发病率数据来自NCCR数据库。我们使用国际疾病分类第10版的标 准,因为只有这版的死亡率数据是可用的。提取的变量有性别、年龄、出生日期、诊断年份、癌症位点、形态学、居住地(乡村和城镇)、地区(中国北部、东北、 东部、中部、南部、西南、西北)。对于分年龄的发病率分析,我们使用了5个大的年龄组(<30,30-44,45-59,60-74,≥75)。

2.3 癌症死亡率数据

我们采用72个登记点编制的数据来估计2015年中国的癌症死亡人数。为了估计2000-2011之间癌症致死的趋势,我们使用了22个登记点 的数据来分析。这些登记点的癌症死亡数据来自地方医院、社区卫生中心、人口统计(包括来自国家疾病监测系统DSP的数据)和民政局。DSP系统由卫生部于 上世纪80年代建立,例行收集由医院提供的死亡认证信息,或者在死亡认证不可用的情况下采用按家走访的方式收集。尽管DSP使用具有全国代表性的位点样 本,但这仍然只覆盖了很小的一部分人口(不到1%)。

2.4 癌症存活率

由于72个登记点没有精确的后续信息,这篇研究中的5年相对存活率使用M/I比率来进行估计,这是一个以前就被使用过的方法。我们只估计整体的 癌症存活率,因为使用这种方法有可能高估或者低估某一种癌症类型的存活率。我们计算M/I比率的前提是,假设2009到2011到2015的发病率和致死 率的关系没有改变,所以我们通过年龄标准化的发病率(2009-2011)来划分年龄标准化的死亡率(2009-2011)。

2.5 人口数据

以5岁年龄组和性别分组的人口数据来源于统计数据或公安人口普查。个人登记提供的数据通过各地区提供给NCCR。这些数据来源于各地局部统计或公安局或基于人口普查数据的计算。

2.6 统计学分析

我们估计了2015年中国的所有癌症的新增病例人数,并根据72个癌症登记点的特定年龄组发病率数据(2009-2011)分性别估计了26个 特定的癌症类型的新增病例数。我们用相同的方法估计了2015的癌症死亡数量。对于10种最普遍的癌症,这些新增患病和死亡的数字还按照城镇/乡村登记点 以及覆盖中国的7个行政区域进行了细分。所有癌症的汇总数据和6个最普遍癌症的数据也按照5个大的年龄组进行了分层 (<30,30-44,45-59,60-74,≥75)。

从2000年到2011年发病率和死亡率的时间趋势分析是通过将连接点模型与对数转化的、年龄标准化的比率来进行计算的。为了减少这段时间内报 道错误改变的可能性,我们将所有的模型限制在最大2个连接点之内。趋势表达为年度变化百分比(APC),我们用Z测试来评估APC是否在统计学上偏离了 0。在描述趋势时,术语“增加”或“减少”用于趋势的斜率(APC)统计学显着时(P<。05)。对于非统计学分析,我们使用术语“稳定”。对于所有的分 析,我们都分性别陈述所有癌症与10种最常见癌症的结果。

  3 、结果

3.1 数据质量

3个主要的指标(MV%, DCO%, M/I比率)对于基于人群的癌症登记、按照癌症类型的分类,表明两种癌症登记数据的质量都很高(图2)。由于22个登记点的半数都通过了IARC的认证, 它们的数据质量被认为高于72个癌症登记点数据,有着更高的MV%和更低的DCO%。这些数据质量参数加上UB%和CPU%都呈现于表1(见在线信息)。

图 2:两套癌症登记数据质量的三个主要测量指标。最左边的一列数字是ICD10 的分类标准。DCO%只表征具有死亡认证的癌症病例百分比。M/I 为死亡率/发病率;MV%为形态学验证的比率。

3.2 2015 预期癌症发病率

预计2015年将会有大约4292000个新增侵袭性癌症病例,与每天12000个新病例相符合。男性中最普遍的5种癌症依次为:肺和支气管癌 症,胃癌,食道癌,肝癌,结直肠癌,这些占到所有癌症病例的三分之二。女性中最普遍的5种癌症依次为:乳腺癌,肺和支气管癌,胃癌,结直肠癌,食道癌,这 些占到了所有癌症病例的60%。单是乳腺癌就占到了所有女性癌症的15%(表2)。

10种最普遍癌症的新增病例数和发病率按照城乡和居住地分组呈现于表3。对于所有的癌症,年龄标准化发病率男性要高于女性 (234.9/168.7每100000人),农村高于城镇(213.6/191.5每100000人)。西南部有最高的癌症发病率,其次为北部和东北; 中部的发病率最低。

3.3 2015 预期癌症死亡率

据估计2015年将会有2814000名中国人死于癌症,与每天7500例癌症死亡相符合。男性和女性死亡率最高的癌症均为:肺和支气管癌、胃 癌、肝癌、食道癌、结直肠癌,占到了所有癌症死亡的四分之三(表2)。与发病率类似,年龄标准化死亡率男性高于女性(165.9/88.8每100000 人),农村高于城市(149.0/109.5每100000人)(表4)。最高的死亡率仍是西南、北部和东北,中部最低。

3.4 2015年分年龄段分性别的发病率和死亡率

在60岁之前,肝癌是被诊断出的最普遍的癌症,并且在男性的癌症死亡中占比最高,其次是肺癌和胃癌,这也是60-74的发病和死亡的主要类型 (表5)。对于75岁以上的男性,肺癌是最广泛被诊断出的癌症,也是最主要的死亡原因。男性的大多数癌症新增病例和死亡都位于60-74岁之间。

在女性中,30岁以前被诊断出的最普遍的癌症是甲状腺癌,30-59岁之间是乳腺癌,60岁以后是肺癌(表5)。45岁以下,乳腺癌是导致死亡的最主要原因,其次是肺癌。60-74岁的新增和死亡病例是最多的。

3.5 2015 年预期癌症存活率

根据预测,所有的癌症结合来看,2015年大概36.9%的癌症患者能够存活5年以上,女性的存活率比男性要好(47.3%/29.3%)(表 6)。根据诊断时的居住地得出的5年存活率估计有一定的潜在变化:农村病人的存活率比城市更低(30.3%/42.8%)。与前面相似,存活率最低的是西 南地区(24.9%),最高的是中部地区(41.0%)。

3.6 癌症发病与死亡的趋势

对于所有的癌症,在研究的时间段内(2000-2011),男性的发病率较为稳定,女性有显著上升(P<.05)(图3,表7)。与此相反,两 性的死亡率都有显著下降(图3,表8)。尽管这个趋势令人高兴,但实际上在此期间癌症的死亡人数增加了(增加了73.8%,从2000年的51090到 2011年的88800),这主要是由于人口增加和老龄化(图4)。

对于男性,在10种最普遍的癌症中,以时间趋势分析,从2000年到2011年发病率增加的有6种,(胰腺癌,结直肠癌,脑和中枢神经系统癌症,前列腺癌,膀胱癌,白血病),而胃癌、食道癌、肝癌则有下降(P<.05)。肺癌的趋势则比较稳定(图5,表7)。

对于女性,10种最普遍的癌症中有6种年龄标准化发病率显著上升(结直肠癌,肺癌,乳腺癌,宫颈癌,子宫体癌,甲状腺癌,P<.05)。与男性相同,胃癌、食道癌、肝癌可见下降趋势(图6,表7)。

对于男性,10种最普遍的癌症中,4种的年龄标准化死亡率可见上升(结直肠癌,胰腺癌,前列腺癌,白血病,P<.05),其他趋势较为稳定(肺癌、膀胱癌和脑癌)(图7,表8)

对于女性,最普遍的10种癌症中3种的死亡率上升(乳腺癌,宫颈癌和卵巢癌),结直肠癌、肺癌、子宫体癌和甲状腺癌趋势较为稳定。(图8,表8)。

与癌症发病率相似,胃癌、食道癌、肝癌的死亡率在两性中都有下降(图7,8)。肺癌的趋势男女都较为稳定,这是两性最主要的癌症死亡原因。

  4.、讨论

4.1 对中国癌症预防的提示

尽管之前已经有过对于全国癌症估计的报道,但这些都只限于特定的年份或癌症种类,很难进行不同癌症间的横向比较。本研究提供了更加全面的全国范围内的癌症统计,使用了最新、最具权威性的数据,包含了信息与时间趋势。

癌症防控需要依赖于基于人群的发病率和死亡率数据,以此来执行政策和评估政策的有效性。因此,最新的全国范围内对于癌症负担和时间趋势的分析, 对理解癌症的病因学,和有效的预防、早期诊断和管理措施有着至关重要的作用。这些结果也会成为未来评估中国癌症控制有效性的基线,也将有助于区分地区间需 求的优先级。

中国人口占到了世界人口的五分之一,因此这些数据将会对世界癌症负担有着重要的作用:大约22%的新增癌

症病例和27%的癌症死亡发生在中国。更重要的是,中国的癌症谱与发达国家明显不同。中国最普遍的4个癌症是肺癌、胃癌、肝癌、食道癌。这几种 癌症占到了中国癌症诊断的57%,而在美国只有18%。同样,中国的这几种癌症占到了全世界发病负担的1/3到1/2。与此相对,美国最普遍的癌症是肺 癌、乳腺癌、前列腺癌和结直肠癌。中国最普遍的癌症生存率很低;而美国的几种除了肺癌之外,预后都非常良好,对于前列腺癌和乳腺癌,有相当的比例是在早期 诊断扫描的时候发现的,因而抬高了发病率。癌症发病类型的差异对于死亡率的差异有着重要的影响。

我们对2015年的预计是基于72个登记点2009-2011年的数据。这些登记点只覆盖了中国人口的6.5%,但它们是目前可用的最好的全国 范围的数据,代表了8550万人口。除此之外,与之前的研究(只覆盖了人口的2%)相比,本研究采用数据的人口覆盖面明显扩大,包含了更多的西部地区,因 而对于中国整体情况就更具有代表性。另外,大陆的全部12个基于人群的癌症登记点都有着高质量的数据,完全满足CI5的标准。全国发病率的估计可以广泛的 与之前发表的进行比较。两个中国之前发表的年度报告估计10年和11年的新增癌症病例分别为309万人和337万人。更早的一个估计是05年的296万 人,尽管使用的方法不同。

我们对于2015年中国发病率(429万人)的估计要明显高于GLOBOCAN在2012年作出的340万人的估计。这些差距主要来源于数据时 间线的不同(09-11/03-07)、代表性的不同、地理覆盖程度的不同(72个登记点覆盖6.5%的人口/23个登记点覆盖3%的人口)。尤其是有着 更高发病率的农村居民(213.6每10万人/191.5每10万人)占到了32.7%,GLOBOCAN2012年的估计只占到了21.5%。用来获取 全国发病率的数据也是不同的,因为GLOBOCAN2012通过把23个登记点的特定年龄、特定性别、特定位置癌症死亡率模型化转化为发病率。需要承认, 这些数据并不都满足IARC的质量标准,已发表报告中的这些差异低估了中国对于提升癌症登记点覆盖率和质量的需求。

与发病率相比,我们的死亡率与GLOBOCAN有着更高的相似性。我们估计2015年中国癌症死亡为281万人,GLOBOCAN2012为 246万人。这反映了两个研究使用的死亡率数据源是相似的:GLOBOCAN用了DSP(04-10),我们使用了72个癌症登记点的数据 (09-11),DSP数据就是这些数据的一部分。DSP数据基于县,按照地理区域划分,被特殊设计为具有全国代表性。对于港澳的发病率和死亡 率,GLOBOCAN2012计入而我们没有计入。

我们发现整体癌症的发病率地区间有显着的差异(农村/城市,地区间)。农村居民比城市居民有更高的发病率,并且发病率在中国的7个行政区域都是 不同的。这些地理差异可能由多种因素造成,但农村更高的吸烟率明显是重要原因。12种由吸烟引起的癌症占到了中国所有癌症的75%。与我们的假设一致,癌 症发病率最高的西南地区在2002年的报道中有着最高的吸烟普遍性。

我们发现癌症的死亡率和存活率的地理差异更大。这些差异或许可以部分解释为,更为有限的医疗资源、更低水平的癌症护理、农村和欠发达地区被诊断出时就已经是晚期的概率更大。这就为政府向农村和欠发达地区投入更多资源和服务提供了理由。

2000-2011之间中国的癌症诊断数量有着显着的上升。很大一部分原因是中国的人口增加和老龄化。其他因素可能也有贡献,比如不健康生活方 式的流行、疾病意识的提升、诊断服务和数据完整性的提升等等。发病率增加幅度最大的是前列腺癌、宫颈癌和女性的甲状腺癌。前列腺癌增加的因素目前还不清 楚;这可能与逐渐应用前列腺特异性抗原扫描和活检水平的提升有关,也有可能是由于逐渐西化的生活方式。西方的生活方式增加了中国的肥胖率,减少了活动量, 可能会对结直肠癌和乳腺癌发病率有影响。乳腺癌发病率的上升也可能与计划生育政策有关。

与发达国家发病率减少的趋势相反,中国的宫颈癌发病率在增加。这可能也反映出了帕帕尼科拉乌试验的不足,中国据报道只有1/5的女性做过帕…… 实验来进行宫颈癌的检查。HPV感染率的增加,和大陆HPV疫苗的不足,表明在可预见的未来,中国的宫颈癌趋势将继续和国际保持差距。

女性甲状腺癌的增加与其他国家相类似;尽管这可能与各种新技术的使用导致的过度诊断有关,但由于缺少疾病阶段的信息,也不能排除发病率确实在增加的可能性。

胃癌、食道癌、肝癌的发病率和死亡率都有明显减少。尽管发病率减少了,但人口基数的增加和老龄化仍然使得新增病例的数量非常大。对感染的控制可能也对这个趋势有影响,比如对导致肝癌的HBV、HCV的控制,对导致

胃癌的幽门螺杆菌的控制。通过对婴儿注射疫苗来进行HBV早期预防已经取得了明显的成效:0-19岁的肝癌死亡率15年内下降了95%。尽管 HBV疫苗的成功对于预防儿童肝癌取得了明显的成效,但也许不能解释对于全年龄段的影响。还有其他很多的重要因素,比如受到黄曲霉毒素污染的玉米的减少和 饮用水的净化。计划生育政策减少了家庭内部的儿童间HBV的传染,更加规范的注射操作减少了医院内HBV和HCV的传染,这些因素可能也会对整体肝癌发病 率有影响。

4.2 对于中国早期诊断和管理的提示

尽管预防措施对于减少长期癌症负担有着重要的作用,但这些措施无法在近期内见效。因此,加强早期诊断和提升医疗服务将是快速缓解中国癌症负担的重要措施。尤其是,地区间的重大差异证明了确保公平的诊断时间、癌症护理可用性和医疗服务质量的重要性。

解决这个问题的一个巨大困难就是中国巨大的人口基数和地理的多样性。即便是按照目前的乳腺癌监测扩张速率,仍需要40年来为目标年龄组的每一个 女性进行一次检测。另外,相比高收入国家更加年轻的诊断患病年龄中位数也为我国提出一些建议,中国的资源应当集中于提高意识和检测乳房肿块时的早期诊断。 尽管有这些地理和人口的障碍,对于食道癌的内窥镜扫描项目正在扩张。另外,新一代基于高风险HPV的扫描测试正逐步应用于中低收入的农村地区。

由于Ⅰ期肺癌的手术治疗已经正式可以显着提升存活率,因此使用低剂量的CT更早地探测肺癌不仅可以降低现有的死亡率,也可以间接提升公共健康预 防和控烟运动的有效性。由于中国很多医院继续使用X光来诊断肺癌,增加医学检测容量,尤其是在农村地区,仍然有着很高的优先级。

为了解决地理多样性和医疗资源分配不均(城市有30%的人口和70%的医疗资源),中国已经实施了癌症护理超级中心策略,集中了很多癌症专家, 处理癌症病例的密度非常大。然而,拿掉得到最优治疗的地理和财政屏障仍然有着最高的优先级,因为农村人口和弱势群体不仅短缺医生,得到医疗服务的距离也更 远。另外,尽管基本医疗保险几乎实现了全覆盖,但它几乎不为癌症治疗负担哪怕部分费用,这就意味着病人要么自费治疗,要么放弃治疗。

任何试图提升早期诊断和治疗的动机都需要考虑中国的独特传统和文化信仰。很多人得了癌症就听天由命,不远谈论治疗和预后,因为无论怎样,得了癌 症就会死。更好的理解这些思想对于实施正确的项目和提升医患间的信任非常重要。与此同时,传统中医已经在中国的卫生体系里面存在了几千年,与中国的文化、 政策、历史都相关。因此,或许可以将癌症的护理和治疗与中医研究中心整合起来。

为了更好地量化早期诊断与治疗对于发病率和死亡率的影响,我们还需要疾病所处阶段和接受治疗的数据。既然这种数据在目前的中国癌症登记体系中是没有的,这就需要具有足够大、有代表性、基于人群的特殊的调查研究。

4.3 限制

尽管这篇研究中的数据覆盖人口是之前研究的两倍,但这仍少于中国人口的十分之一。仍然有许多未知水平的不确定因素。尽管我们在控制数据质量方面 已经付出了很大的努力,但数据质量中仍然有许多变数。M/I比率被用作近五年相对存活率的近似值,这种解释可能是有问题的,因为死亡率和发病率可能是涉及 到完全不同的人群。这使得它更容易受发病率的影响,因此为更容易死亡的癌症提供了更加精确的估计。然而,17个登记点的未发布数据表明,所有癌症的M/I 比率只比计算得到的5年相对存活率高了1.4%。最终,对于一个14亿人的国家来说,要保证分子面临的风险和分母相同实在是一个不小的挑战,尤其是考虑到 在大城市医疗机构里面治疗的病例和来自农村的移民病例。患病案例的地理信息是基于永久居住地而不是治疗地。另外,通过城居和新农合得到的外出务工者(占人 口的9%)的癌症诊断,都是基于他们的户口登记得到的。

  5 、结论

为了制定一个恰当的癌症控制计划,拥有一套细节的、有代表性的、精确的、基于人群的数据是非常关键的。这些评估和癌症登记的努力都是为了达成这 一目标所进行的重要步骤。尽管这些全国估算中仍然有不精确的地方,但这都是基于可用的最优数据来进行的发病率和死亡率的估计。这可能为中国未来的癌症防控 提供可供比较的基线和评价标准,并帮助发现最需要援助的地区。根据国际经验,当需求更加明确、有更详实的证据支持时,政府和其他卫生服务提供者将会更有动 力提供帮助。根据这篇研究的数据,中国正面临,并且未来将继续面临极大的癌症压力,因此需要政府和各非政府组织的共同努力。

关键区域可能是总体水平上临床癌症护理水平的提升,通过有目标的政策改革和投资来提升农村地区的医疗服务水平,为弱势群体提供医疗服务。癌症的 初级预防项目,比如控烟和缓和西式生活方式的不良影响,提高早期诊断的有效性和覆盖率,这些对于逆转中国癌症的流行趋势至关重要。保证现有的空气和水污染 控制法律得到有效实行仍是当务之急。考虑到中国对世界癌症负担的重要性,特别是4 种主要癌症(肺癌、肝癌、胃癌、食道癌),我们必须采取适当的策略和政策来减少这些可预防的癌症(通过减少烟草的流行和与癌症相关的感染),这将对中国和 世界的癌症负担有着重要的影响。

Deep Learning System Improves Breast Cancer Detection

Researchers from Beth Israel Deaconess Medical Center (BIDMC) and Harvard Medical School have developed a deep learning approach to read and interpret pathology images.

Trained on Tesla K80 GPUs with the cuDNN-accelerated Caffe deep learning framework, their system achieved 92 percent accuracy at identifying breast cancer in images of lymph nodes which earned them the top prize in two separate categories at the annual International Symposium of Biomedical Imaging (ISBI) challenge. The team also published a paper detailing more of their work.

For the slide-based classification task, human pathologists were accurate 96 percent of the time.

DL Breast Cancer Detection Image
The framework used for breast cancer detection.

Andrew Beck from BIDMC said what’s truly exciting is that 99.5 percent accuracy can be achieved when the pathologists’ analysis and results from the deep learning system are used together. He added, “Our results in the ISBI competition show that what the computer is doing is genuinely intelligent and that the combination of human and computer interpretations will result in more precise and more clinically valuable diagnoses to guide treatment decisions.”

Awesome Machine Learning

Table of Contents

APL

General-Purpose Machine Learning

  • naive-apl – Naive Bayesian Classifier implementation in APL

C

General-Purpose Machine Learning

  • Recommender – A C library for product recommendations/suggestions using collaborative filtering (CF).
  • Darknet – Darknet is an open source neural network framework written in C and CUDA. It is fast, easy to install, and supports CPU and GPU computation.

Computer Vision

  • CCV – C-based/Cached/Core Computer Vision Library, A Modern Computer Vision Library
  • VLFeat – VLFeat is an open and portable library of computer vision algorithms, which has Matlab toolbox

Speech Recognition

  • HTK -The Hidden Markov Model Toolkit (HTK) is a portable toolkit for building and manipulating hidden Markov models.

C++

Computer Vision

  • OpenCV – OpenCV has C++, C, Python, Java and MATLAB interfaces and supports Windows, Linux, Android and Mac OS.
  • DLib – DLib has C++ and Python interfaces for face detection and training general object detectors.
  • EBLearn – Eblearn is an object-oriented C++ library that implements various machine learning models
  • VIGRA – VIGRA is a generic cross-platform C++ computer vision and machine learning library for volumes of arbitrary dimensionality with Python bindings.

General-Purpose Machine Learning

  • mlpack – A scalable C++ machine learning library
  • DLib – A suite of ML tools designed to be easy to imbed in other applications
  • encog-cpp
  • shark
  • Vowpal Wabbit (VW) – A fast out-of-core learning system.
  • sofia-ml – Suite of fast incremental algorithms.
  • Shogun – The Shogun Machine Learning Toolbox
  • Caffe – A deep learning framework developed with cleanliness, readability, and speed in mind. [DEEP LEARNING]
  • CXXNET – Yet another deep learning framework with less than 1000 lines core code [DEEP LEARNING]
  • XGBoost – A parallelized optimized general purpose gradient boosting library.
  • CUDA – This is a fast C++/CUDA implementation of convolutional [DEEP LEARNING]
  • Stan – A probabilistic programming language implementing full Bayesian statistical inference with Hamiltonian Monte Carlo sampling
  • BanditLib – A simple Multi-armed Bandit library.
  • Timbl – A software package/C++ library implementing several memory-based learning algorithms, among which IB1-IG, an implementation of k-nearest neighbor classification, and IGTree, a decision-tree approximation of IB1-IG. Commonly used for NLP.
  • Disrtibuted Machine learning Tool Kit (DMTK) – A distributed machine learning (parameter server) framework by Microsoft. Enables training models on large data sets across multiple machines. Current tools bundled with it include: LightLDA and Distributed (Multisense) Word Embedding.
  • igraph – General purpose graph library
  • Warp-CTC – A fast parallel implementation of Connectionist Temporal Classification (CTC), on both CPU and GPU.
  • CNTK – The Computational Network Toolkit (CNTK) by Microsoft Research, is a unified deep-learning toolkit that describes neural networks as a series of computational steps via a directed graph.
  • DeepDetect – A machine learning API and server written in C++11. It makes state of the art machine learning easy to work with and integrate into existing applications.
  • Fido – A highly-modular C++ machine learning library for embedded electronics and robotics.
  • DSSTNE – A software library created by Amazon for training and deploying deep neural networks using GPUs which emphasizes speed and scale over experimental flexibility.
  • Intel(R) DAAL – A high performance software library developed by Intel and optimized for Intel’s architectures. Library provides algorithmic building blocks for all stages of data analytics and allows to process data in batch, online and distributed modes.

Natural Language Processing

  • MIT Information Extraction Toolkit – C, C++, and Python tools for named entity recognition and relation extraction
  • CRF++ – Open source implementation of Conditional Random Fields (CRFs) for segmenting/labeling sequential data & other Natural Language Processing tasks.
  • CRFsuite – CRFsuite is an implementation of Conditional Random Fields (CRFs) for labeling sequential data.
  • BLLIP Parser – BLLIP Natural Language Parser (also known as the Charniak-Johnson parser)
  • colibri-core – C++ library, command line tools, and Python binding for extracting and working with basic linguistic constructions such as n-grams and skipgrams in a quick and memory-efficient way.
  • ucto – Unicode-aware regular-expression based tokenizer for various languages. Tool and C++ library. Supports FoLiA format.
  • libfolia – C++ library for the FoLiA format
  • frog – Memory-based NLP suite developed for Dutch: PoS tagger, lemmatiser, dependency parser, NER, shallow parser, morphological analyzer.
  • MeTAMeTA : ModErn Text Analysis is a C++ Data Sciences Toolkit that facilitates mining big text data.

Speech Recognition

  • Kaldi – Kaldi is a toolkit for speech recognition written in C++ and licensed under the Apache License v2.0. Kaldi is intended for use by speech recognition researchers.

Sequence Analysis

  • ToPS – This is an objected-oriented framework that facilitates the integration of probabilistic models for sequences over a user defined alphabet.

Gesture Detection

  • grt – The Gesture Recognition Toolkit (GRT) is a cross-platform, open-source, C++ machine learning library designed for real-time gesture recognition.

Common Lisp

General-Purpose Machine Learning

  • mgl – Neural networks (boltzmann machines, feed-forward and recurrent nets), Gaussian Processes
  • mgl-gpr – Evolutionary algorithms
  • cl-libsvm – Wrapper for the libsvm support vector machine library

Clojure

Natural Language Processing

  • Clojure-openNLP – Natural Language Processing in Clojure (opennlp)
  • Infections-clj – Rails-like inflection library for Clojure and ClojureScript

General-Purpose Machine Learning

  • Touchstone – Clojure A/B testing library
  • Clojush – The Push programming language and the PushGP genetic programming system implemented in Clojure
  • Infer – Inference and machine learning in clojure
  • Clj-ML – A machine learning library for Clojure built on top of Weka and friends
  • Encog – Clojure wrapper for Encog (v3) (Machine-Learning framework that specializes in neural-nets)
  • Fungp – A genetic programming library for Clojure
  • Statistiker – Basic Machine Learning algorithms in Clojure.
  • clortex – General Machine Learning library using Numenta’s Cortical Learning Algorithm
  • comportex – Functionally composable Machine Learning library using Numenta’s Cortical Learning Algorithm

Data Analysis / Data Visualization

  • Incanter – Incanter is a Clojure-based, R-like platform for statistical computing and graphics.
  • PigPen – Map-Reduce for Clojure.
  • Envision – Clojure Data Visualisation library, based on Statistiker and D3

Elixir

General-Purpose Machine Learning

  • Simple Bayes – A Simple Bayes / Naive Bayes implementation in Elixir.

Natural Language Processing

  • Stemmer – An English (Porter2) stemming implementation in Elixir.

Erlang

General-Purpose Machine Learning

  • Disco – Map Reduce in Erlang

Go

Natural Language Processing

  • go-porterstemmer – A native Go clean room implementation of the Porter Stemming algorithm.
  • paicehusk – Golang implementation of the Paice/Husk Stemming Algorithm.
  • snowball – Snowball Stemmer for Go.
  • go-ngram – In-memory n-gram index with compression.

General-Purpose Machine Learning

  • gago – Multi-population, flexible, parallel genetic algorithm.
  • Go Learn – Machine Learning for Go
  • go-pr – Pattern recognition package in Go lang.
  • go-ml – Linear / Logistic regression, Neural Networks, Collaborative Filtering and Gaussian Multivariate Distribution
  • bayesian – Naive Bayesian Classification for Golang.
  • go-galib – Genetic Algorithms library written in Go / golang
  • Cloudforest – Ensembles of decision trees in go/golang.
  • gobrain – Neural Networks written in go
  • GoNN – GoNN is an implementation of Neural Network in Go Language, which includes BPNN, RBF, PCN
  • MXNet – Lightweight, Portable, Flexible Distributed/Mobile Deep Learning with Dynamic, Mutation-aware Dataflow Dep Scheduler; for Python, R, Julia, Go, Javascript and more.

Data Analysis / Data Visualization

  • go-graph – Graph library for Go/golang language.
  • SVGo – The Go Language library for SVG generation
  • RF – Random forests implementation in Go

Haskell

General-Purpose Machine Learning

  • haskell-ml – Haskell implementations of various ML algorithms.
  • HLearn – a suite of libraries for interpreting machine learning models according to their algebraic structure.
  • hnn – Haskell Neural Network library.
  • hopfield-networks – Hopfield Networks for unsupervised learning in Haskell.
  • caffegraph – A DSL for deep neural networks
  • LambdaNet – Configurable Neural Networks in Haskell

Java

Natural Language Processing

  • Cortical.io – Retina: an API performing complex NLP operations (disambiguation, classification, streaming text filtering, etc…) as quickly and intuitively as the brain.
  • CoreNLP – Stanford CoreNLP provides a set of natural language analysis tools which can take raw English language text input and give the base forms of words
  • Stanford Parser – A natural language parser is a program that works out the grammatical structure of sentences
  • Stanford POS Tagger – A Part-Of-Speech Tagger (POS Tagger
  • Stanford Name Entity Recognizer – Stanford NER is a Java implementation of a Named Entity Recognizer.
  • Stanford Word Segmenter – Tokenization of raw text is a standard pre-processing step for many NLP tasks.
  • Tregex, Tsurgeon and Semgrex – Tregex is a utility for matching patterns in trees, based on tree relationships and regular expression matches on nodes (the name is short for “tree regular expressions”).
  • Stanford Phrasal: A Phrase-Based Translation System
  • Stanford English Tokenizer – Stanford Phrasal is a state-of-the-art statistical phrase-based machine translation system, written in Java.
  • Stanford Tokens Regex – A tokenizer divides text into a sequence of tokens, which roughly correspond to “words”
  • Stanford Temporal Tagger – SUTime is a library for recognizing and normalizing time expressions.
  • Stanford SPIED – Learning entities from unlabeled text starting with seed sets using patterns in an iterative fashion
  • Stanford Topic Modeling Toolbox – Topic modeling tools to social scientists and others who wish to perform analysis on datasets
  • Twitter Text Java – A Java implementation of Twitter’s text processing library
  • MALLET – A Java-based package for statistical natural language processing, document classification, clustering, topic modeling, information extraction, and other machine learning applications to text.
  • OpenNLP – a machine learning based toolkit for the processing of natural language text.
  • LingPipe – A tool kit for processing text using computational linguistics.
  • ClearTK – ClearTK provides a framework for developing statistical natural language processing (NLP) components in Java and is built on top of Apache UIMA.
  • Apache cTAKES – Apache clinical Text Analysis and Knowledge Extraction System (cTAKES) is an open-source natural language processing system for information extraction from electronic medical record clinical free-text.
  • ClearNLP – The ClearNLP project provides software and resources for natural language processing. The project started at the Center for Computational Language and EducAtion Research, and is currently developed by the Center for Language and Information Research at Emory University. This project is under the Apache 2 license.
  • CogcompNLP – This project collects a number of core libraries for Natural Language Processing (NLP) developed in the University of Illinois’ Cognitive Computation Group, for example illinois-core-utilities which provides a set of NLP-friendly data structures and a number of NLP-related utilities that support writing NLP applications, running experiments, etc, illinois-edison a library for feature extraction from illinois-core-utilities data structures and many other packages.

General-Purpose Machine Learning

  • aerosolve – A machine learning library by Airbnb designed from the ground up to be human friendly.
  • Datumbox – Machine Learning framework for rapid development of Machine Learning and Statistical applications
  • ELKI – Java toolkit for data mining. (unsupervised: clustering, outlier detection etc.)
  • Encog – An advanced neural network and machine learning framework. Encog contains classes to create a wide variety of networks, as well as support classes to normalize and process data for these neural networks. Encog trains using multithreaded resilient propagation. Encog can also make use of a GPU to further speed processing time. A GUI based workbench is also provided to help model and train neural networks.
  • FlinkML in Apache Flink – Distributed machine learning library in Flink
  • H2O – ML engine that supports distributed learning on Hadoop, Spark or your laptop via APIs in R, Python, Scala, REST/JSON.
  • htm.java – General Machine Learning library using Numenta’s Cortical Learning Algorithm
  • java-deeplearning – Distributed Deep Learning Platform for Java, Clojure,Scala
  • Mahout – Distributed machine learning
  • Meka – An open source implementation of methods for multi-label classification and evaluation (extension to Weka).
  • MLlib in Apache Spark – Distributed machine learning library in Spark
  • Neuroph – Neuroph is lightweight Java neural network framework
  • ORYX – Lambda Architecture Framework using Apache Spark and Apache Kafka with a specialization for real-time large-scale machine learning.
  • Samoa SAMOA is a framework that includes distributed machine learning for data streams with an interface to plug-in different stream processing platforms.
  • RankLib – RankLib is a library of learning to rank algorithms
  • rapaio – statistics, data mining and machine learning toolbox in Java
  • RapidMiner – RapidMiner integration into Java code
  • Stanford Classifier – A classifier is a machine learning tool that will take data items and place them into one of k classes.
  • SmileMiner – Statistical Machine Intelligence & Learning Engine
  • SystemML – flexible, scalable machine learning (ML) language.
  • WalnutiQ – object oriented model of the human brain
  • Weka – Weka is a collection of machine learning algorithms for data mining tasks
  • LBJava – Learning Based Java is a modeling language for the rapid development of software systems, offers a convenient, declarative syntax for classifier and constraint definition directly in terms of the objects in the programmer’s application.

Speech Recognition

  • CMU Sphinx – Open Source Toolkit For Speech Recognition purely based on Java speech recognition library.

Data Analysis / Data Visualization

  • Flink – Open source platform for distributed stream and batch data processing.
  • Hadoop – Hadoop/HDFS
  • Spark – Spark is a fast and general engine for large-scale data processing.
  • Storm – Storm is a distributed realtime computation system.
  • Impala – Real-time Query for Hadoop
  • DataMelt – Mathematics software for numeric computation, statistics, symbolic calculations, data analysis and data visualization.
  • Dr. Michael Thomas Flanagan’s Java Scientific Library

Deep Learning

  • Deeplearning4j – Scalable deep learning for industry with parallel GPUs

Javascript

Natural Language Processing

  • Twitter-text – A JavaScript implementation of Twitter’s text processing library
  • NLP.js – NLP utilities in javascript and coffeescript
  • natural – General natural language facilities for node
  • Knwl.js – A Natural Language Processor in JS
  • Retext – Extensible system for analyzing and manipulating natural language
  • TextProcessing – Sentiment analysis, stemming and lemmatization, part-of-speech tagging and chunking, phrase extraction and named entity recognition.
  • NLP Compromise – Natural Language processing in the browser

Data Analysis / Data Visualization

  • D3.js
  • High Charts
  • NVD3.js
  • dc.js
  • chartjs
  • dimple
  • amCharts
  • D3xter – Straight forward plotting built on D3
  • statkit – Statistics kit for JavaScript
  • datakit – A lightweight framework for data analysis in JavaScript
  • science.js – Scientific and statistical computing in JavaScript.
  • Z3d – Easily make interactive 3d plots built on Three.js
  • Sigma.js – JavaScript library dedicated to graph drawing.
  • C3.js– customizable library based on D3.js for easy chart drawing.
  • Datamaps– Customizable SVG map/geo visualizations using D3.js.
  • ZingChart– library written on Vanilla JS for big data visualization.
  • cheminfo – Platform for data visualization and analysis, using the visualizer project.

General-Purpose Machine Learning

  • Convnet.js – ConvNetJS is a Javascript library for training Deep Learning models[DEEP LEARNING]
  • Clusterfck – Agglomerative hierarchical clustering implemented in Javascript for Node.js and the browser
  • Clustering.js – Clustering algorithms implemented in Javascript for Node.js and the browser
  • Decision Trees – NodeJS Implementation of Decision Tree using ID3 Algorithm
  • DN2A – Digital Neural Networks Architecture
  • figue – K-means, fuzzy c-means and agglomerative clustering
  • Node-fann – FANN (Fast Artificial Neural Network Library) bindings for Node.js
  • Kmeans.js – Simple Javascript implementation of the k-means algorithm, for node.js and the browser
  • LDA.js – LDA topic modeling for node.js
  • Learning.js – Javascript implementation of logistic regression/c4.5 decision tree
  • Machine Learning – Machine learning library for Node.js
  • mil-tokyo – List of several machine learning libraries
  • Node-SVM – Support Vector Machine for nodejs
  • Brain – Neural networks in JavaScript [Deprecated]
  • Bayesian-Bandit – Bayesian bandit implementation for Node and the browser.
  • Synaptic – Architecture-free neural network library for node.js and the browser
  • kNear – JavaScript implementation of the k nearest neighbors algorithm for supervised learning
  • NeuralN – C++ Neural Network library for Node.js. It has advantage on large dataset and multi-threaded training.
  • kalman – Kalman filter for Javascript.
  • shaman – node.js library with support for both simple and multiple linear regression.
  • ml.js – Machine learning and numerical analysis tools for Node.js and the Browser!
  • Pavlov.js – Reinforcement learning using Markov Decision Processes
  • MXNet – Lightweight, Portable, Flexible Distributed/Mobile Deep Learning with Dynamic, Mutation-aware Dataflow Dep Scheduler; for Python, R, Julia, Go, Javascript and more.

Misc

  • sylvester – Vector and Matrix math for JavaScript.
  • simple-statistics – A JavaScript implementation of descriptive, regression, and inference statistics. Implemented in literate JavaScript with no dependencies, designed to work in all modern browsers (including IE) as well as in node.js.
  • regression-js – A javascript library containing a collection of least squares fitting methods for finding a trend in a set of data.
  • Lyric – Linear Regression library.
  • GreatCircle – Library for calculating great circle distance.

Julia

General-Purpose Machine Learning

  • MachineLearning – Julia Machine Learning library
  • MLBase – A set of functions to support the development of machine learning algorithms
  • PGM – A Julia framework for probabilistic graphical models.
  • DA – Julia package for Regularized Discriminant Analysis
  • Regression – Algorithms for regression analysis (e.g. linear regression and logistic regression)
  • Local Regression – Local regression, so smooooth!
  • Naive Bayes – Simple Naive Bayes implementation in Julia
  • Mixed Models – A Julia package for fitting (statistical) mixed-effects models
  • Simple MCMC – basic mcmc sampler implemented in Julia
  • Distance – Julia module for Distance evaluation
  • Decision Tree – Decision Tree Classifier and Regressor
  • Neural – A neural network in Julia
  • MCMC – MCMC tools for Julia
  • Mamba – Markov chain Monte Carlo (MCMC) for Bayesian analysis in Julia
  • GLM – Generalized linear models in Julia
  • Online Learning
  • GLMNet – Julia wrapper for fitting Lasso/ElasticNet GLM models using glmnet
  • Clustering – Basic functions for clustering data: k-means, dp-means, etc.
  • SVM – SVM’s for Julia
  • Kernal Density – Kernel density estimators for julia
  • Dimensionality Reduction – Methods for dimensionality reduction
  • NMF – A Julia package for non-negative matrix factorization
  • ANN – Julia artificial neural networks
  • Mocha – Deep Learning framework for Julia inspired by Caffe
  • XGBoost – eXtreme Gradient Boosting Package in Julia
  • ManifoldLearning – A Julia package for manifold learning and nonlinear dimensionality reduction
  • MXNet – Lightweight, Portable, Flexible Distributed/Mobile Deep Learning with Dynamic, Mutation-aware Dataflow Dep Scheduler; for Python, R, Julia, Go, Javascript and more.
  • Merlin – Flexible Deep Learning Framework in Julia
  • ROCAnalysis – Receiver Operating Characteristics and functions for evaluation probabilistic binary classifiers
  • GaussianMixtures – Large scale Gaussian Mixture Models
  • ScikitLearn – Julia implementation of the scikit-learn API

Natural Language Processing

Data Analysis / Data Visualization

  • Graph Layout – Graph layout algorithms in pure Julia
  • Data Frames Meta – Metaprogramming tools for DataFrames
  • Julia Data – library for working with tabular data in Julia
  • Data Read – Read files from Stata, SAS, and SPSS
  • Hypothesis Tests – Hypothesis tests for Julia
  • Gadfly – Crafty statistical graphics for Julia.
  • Stats – Statistical tests for Julia
  • RDataSets – Julia package for loading many of the data sets available in R
  • DataFrames – library for working with tabular data in Julia
  • Distributions – A Julia package for probability distributions and associated functions.
  • Data Arrays – Data structures that allow missing values
  • Time Series – Time series toolkit for Julia
  • Sampling – Basic sampling algorithms for Julia

Misc Stuff / Presentations

  • DSP – Digital Signal Processing (filtering, periodograms, spectrograms, window functions).
  • JuliaCon Presentations – Presentations for JuliaCon
  • SignalProcessing – Signal Processing tools for Julia
  • Images – An image library for Julia

Lua

General-Purpose Machine Learning

  • Torch7
    • cephes – Cephes mathematical functions library, wrapped for Torch. Provides and wraps the 180+ special mathematical functions from the Cephes mathematical library, developed by Stephen L. Moshier. It is used, among many other places, at the heart of SciPy.
    • autograd – Autograd automatically differentiates native Torch code. Inspired by the original Python version.
    • graph – Graph package for Torch
    • randomkit – Numpy’s randomkit, wrapped for Torch
    • signal – A signal processing toolbox for Torch-7. FFT, DCT, Hilbert, cepstrums, stft
    • nn – Neural Network package for Torch
    • torchnet – framework for torch which provides a set of abstractions aiming at encouraging code re-use as well as encouraging modular programming
    • nngraph – This package provides graphical computation for nn library in Torch7.
    • nnx – A completely unstable and experimental package that extends Torch’s builtin nn library
    • rnn – A Recurrent Neural Network library that extends Torch’s nn. RNNs, LSTMs, GRUs, BRNNs, BLSTMs, etc.
    • dpnn – Many useful features that aren’t part of the main nn package.
    • dp – A deep learning library designed for streamlining research and development using the Torch7 distribution. It emphasizes flexibility through the elegant use of object-oriented design patterns.
    • optim – An optimization library for Torch. SGD, Adagrad, Conjugate-Gradient, LBFGS, RProp and more.
    • unsup – A package for unsupervised learning in Torch. Provides modules that are compatible with nn (LinearPsd, ConvPsd, AutoEncoder, …), and self-contained algorithms (k-means, PCA).
    • manifold – A package to manipulate manifolds
    • svm – Torch-SVM library
    • lbfgs – FFI Wrapper for liblbfgs
    • vowpalwabbit – An old vowpalwabbit interface to torch.
    • OpenGM – OpenGM is a C++ library for graphical modeling, and inference. The Lua bindings provide a simple way of describing graphs, from Lua, and then optimizing them with OpenGM.
    • sphagetti – Spaghetti (sparse linear) module for torch7 by @MichaelMathieu
    • LuaSHKit – A lua wrapper around the Locality sensitive hashing library SHKit
    • kernel smoothing – KNN, kernel-weighted average, local linear regression smoothers
    • cutorch – Torch CUDA Implementation
    • cunn – Torch CUDA Neural Network Implementation
    • imgraph – An image/graph library for Torch. This package provides routines to construct graphs on images, segment them, build trees out of them, and convert them back to images.
    • videograph – A video/graph library for Torch. This package provides routines to construct graphs on videos, segment them, build trees out of them, and convert them back to videos.
    • saliency – code and tools around integral images. A library for finding interest points based on fast integral histograms.
    • stitch – allows us to use hugin to stitch images and apply same stitching to a video sequence
    • sfm – A bundle adjustment/structure from motion package
    • fex – A package for feature extraction in Torch. Provides SIFT and dSIFT modules.
    • OverFeat – A state-of-the-art generic dense feature extractor
  • Numeric Lua
  • Lunatic Python
  • SciLua
  • Lua – Numerical Algorithms
  • Lunum

Demos and Scripts

  • Core torch7 demos repository.
    • linear-regression, logistic-regression
    • face detector (training and detection as separate demos)
    • mst-based-segmenter
    • train-a-digit-classifier
    • train-autoencoder
    • optical flow demo
    • train-on-housenumbers
    • train-on-cifar
    • tracking with deep nets
    • kinect demo
    • filter-bank visualization
    • saliency-networks
  • Training a Convnet for the Galaxy-Zoo Kaggle challenge(CUDA demo)
  • Music Tagging – Music Tagging scripts for torch7
  • torch-datasets – Scripts to load several popular datasets including:
    • BSR 500
    • CIFAR-10
    • COIL
    • Street View House Numbers
    • MNIST
    • NORB
  • Atari2600 – Scripts to generate a dataset with static frames from the Arcade Learning Environment

Matlab

Computer Vision

  • Contourlets – MATLAB source code that implements the contourlet transform and its utility functions.
  • Shearlets – MATLAB code for shearlet transform
  • Curvelets – The Curvelet transform is a higher dimensional generalization of the Wavelet transform designed to represent images at different scales and different angles.
  • Bandlets – MATLAB code for bandlet transform
  • mexopencv – Collection and a development kit of MATLAB mex functions for OpenCV library

Natural Language Processing

  • NLP – An NLP library for Matlab

General-Purpose Machine Learning

Data Analysis / Data Visualization

  • matlab_gbl – MatlabBGL is a Matlab package for working with graphs.
  • gamic – Efficient pure-Matlab implementations of graph algorithms to complement MatlabBGL’s mex functions.

.NET

Computer Vision

  • OpenCVDotNet – A wrapper for the OpenCV project to be used with .NET applications.
  • Emgu CV – Cross platform wrapper of OpenCV which can be compiled in Mono to e run on Windows, Linus, Mac OS X, iOS, and Android.
  • AForge.NET – Open source C# framework for developers and researchers in the fields of Computer Vision and Artificial Intelligence. Development has now shifted to GitHub.
  • Accord.NET – Together with AForge.NET, this library can provide image processing and computer vision algorithms to Windows, Windows RT and Windows Phone. Some components are also available for Java and Android.

Natural Language Processing

  • Stanford.NLP for .NET – A full port of Stanford NLP packages to .NET and also available precompiled as a NuGet package.

General-Purpose Machine Learning

  • Accord-Framework -The Accord.NET Framework is a complete framework for building machine learning, computer vision, computer audition, signal processing and statistical applications.
  • Accord.MachineLearning – Support Vector Machines, Decision Trees, Naive Bayesian models, K-means, Gaussian Mixture models and general algorithms such as Ransac, Cross-validation and Grid-Search for machine-learning applications. This package is part of the Accord.NET Framework.
  • DiffSharp – An automatic differentiation (AD) library providing exact and efficient derivatives (gradients, Hessians, Jacobians, directional derivatives, and matrix-free Hessian- and Jacobian-vector products) for machine learning and optimization applications. Operations can be nested to any level, meaning that you can compute exact higher-order derivatives and differentiate functions that are internally making use of differentiation, for applications such as hyperparameter optimization.
  • Vulpes – Deep belief and deep learning implementation written in F# and leverages CUDA GPU execution with Alea.cuBase.
  • Encog – An advanced neural network and machine learning framework. Encog contains classes to create a wide variety of networks, as well as support classes to normalize and process data for these neural networks. Encog trains using multithreaded resilient propagation. Encog can also make use of a GPU to further speed processing time. A GUI based workbench is also provided to help model and train neural networks.
  • Neural Network Designer – DBMS management system and designer for neural networks. The designer application is developed using WPF, and is a user interface which allows you to design your neural network, query the network, create and configure chat bots that are capable of asking questions and learning from your feed back. The chat bots can even scrape the internet for information to return in their output as well as to use for learning.

Data Analysis / Data Visualization

  • numl – numl is a machine learning library intended to ease the use of using standard modeling techniques for both prediction and clustering.
  • Math.NET Numerics – Numerical foundation of the Math.NET project, aiming to provide methods and algorithms for numerical computations in science, engineering and every day use. Supports .Net 4.0, .Net 3.5 and Mono on Windows, Linux and Mac; Silverlight 5, WindowsPhone/SL 8, WindowsPhone 8.1 and Windows 8 with PCL Portable Profiles 47 and 344; Android/iOS with Xamarin.
  • Sho – Sho is an interactive environment for data analysis and scientific computing that lets you seamlessly connect scripts (in IronPython) with compiled code (in .NET) to enable fast and flexible prototyping. The environment includes powerful and efficient libraries for linear algebra as well as data visualization that can be used from any .NET language, as well as a feature-rich interactive shell for rapid development.

Objective C

General-Purpose Machine Learning

  • YCML – A Machine Learning framework for Objective-C and Swift (OS X / iOS).
  • MLPNeuralNet – Fast multilayer perceptron neural network library for iOS and Mac OS X. MLPNeuralNet predicts new examples by trained neural network. It is built on top of the Apple’s Accelerate Framework, using vectorized operations and hardware acceleration if available.
  • MAChineLearning – An Objective-C multilayer perceptron library, with full support for training through backpropagation. Implemented using vDSP and vecLib, it’s 20 times faster than its Java equivalent. Includes sample code for use from Swift.
  • BPN-NeuralNetwork – It implemented 3 layers neural network ( Input Layer, Hidden Layer and Output Layer ) and it named Back Propagation Neural Network (BPN). This network can be used in products recommendation, user behavior analysis, data mining and data analysis.
  • Multi-Perceptron-NeuralNetwork – it implemented multi-perceptrons neural network (ニューラルネットワーク) based on Back Propagation Neural Network (BPN) and designed unlimited-hidden-layers.
  • KRHebbian-Algorithm – It is a non-supervisor and self-learning algorithm (adjust the weights) in neural network of Machine Learning.
  • KRKmeans-Algorithm – It implemented K-Means the clustering and classification algorithm. It could be used in data mining and image compression.
  • KRFuzzyCMeans-Algorithm – It implemented Fuzzy C-Means (FCM) the fuzzy clustering / classification algorithm on Machine Learning. It could be used in data mining and image compression.

OCaml

General-Purpose Machine Learning

  • Oml – A general statistics and machine learning library.
  • GPR – Efficient Gaussian Process Regression in OCaml.
  • Libra-Tk – Algorithms for learning and inference with discrete probabilistic models.

PHP

Natural Language Processing

  • jieba-php – Chinese Words Segmentation Utilities.

General-Purpose Machine Learning

  • PredictionBuilder – A library for machine learning that builds predictions using a linear regression.

Python

Computer Vision

  • Scikit-Image – A collection of algorithms for image processing in Python.
  • SimpleCV – An open source computer vision framework that gives access to several high-powered computer vision libraries, such as OpenCV. Written on Python and runs on Mac, Windows, and Ubuntu Linux.
  • Vigranumpy – Python bindings for the VIGRA C++ computer vision library.
  • OpenFace – Free and open source face recognition with deep neural networks.
  • PCV – Open source Python module for computer vision

Natural Language Processing

  • NLTK – A leading platform for building Python programs to work with human language data.
  • Pattern – A web mining module for the Python programming language. It has tools for natural language processing, machine learning, among others.
  • Quepy – A python framework to transform natural language questions to queries in a database query language
  • TextBlob – Providing a consistent API for diving into common natural language processing (NLP) tasks. Stands on the giant shoulders of NLTK and Pattern, and plays nicely with both.
  • YAlign – A sentence aligner, a friendly tool for extracting parallel sentences from comparable corpora.
  • jieba – Chinese Words Segmentation Utilities.
  • SnowNLP – A library for processing Chinese text.
  • spammy – A library for email Spam filtering built on top of nltk
  • loso – Another Chinese segmentation library.
  • genius – A Chinese segment base on Conditional Random Field.
  • KoNLPy – A Python package for Korean natural language processing.
  • nut – Natural language Understanding Toolkit
  • Rosetta – Text processing tools and wrappers (e.g. Vowpal Wabbit)
  • BLLIP Parser – Python bindings for the BLLIP Natural Language Parser (also known as the Charniak-Johnson parser)
  • PyNLPl – Python Natural Language Processing Library. General purpose NLP library for Python. Also contains some specific modules for parsing common NLP formats, most notably for FoLiA, but also ARPA language models, Moses phrasetables, GIZA++ alignments.
  • python-ucto – Python binding to ucto (a unicode-aware rule-based tokenizer for various languages)
  • python-frog – Python binding to Frog, an NLP suite for Dutch. (pos tagging, lemmatisation, dependency parsing, NER)
  • python-zpar – Python bindings for ZPar, a statistical part-of-speech-tagger, constiuency parser, and dependency parser for English.
  • colibri-core – Python binding to C++ library for extracting and working with with basic linguistic constructions such as n-grams and skipgrams in a quick and memory-efficient way.
  • spaCy – Industrial strength NLP with Python and Cython.
  • PyStanfordDependencies – Python interface for converting Penn Treebank trees to Stanford Dependencies.
  • Distance – Levenshtein and Hamming distance computation
  • Fuzzy Wuzzy – Fuzzy String Matching in Python
  • jellyfish – a python library for doing approximate and phonetic matching of strings.
  • editdistance – fast implementation of edit distance
  • textacy – higher-level NLP built on Spacy

General-Purpose Machine Learning

  • machine learning – automated build consisting of a web-interface, and set of programmatic-interface API, for support vector machines. Corresponding dataset(s) are stored into a SQL database, then generated model(s) used for prediction(s), are stored into a NoSQL datastore.
  • XGBoost – Python bindings for eXtreme Gradient Boosting (Tree) Library
  • Bayesian Methods for Hackers – Book/iPython notebooks on Probabilistic Programming in Python
  • Featureforge A set of tools for creating and testing machine learning features, with a scikit-learn compatible API
  • MLlib in Apache Spark – Distributed machine learning library in Spark
  • scikit-learn – A Python module for machine learning built on top of SciPy.
  • metric-learn – A Python module for metric learning.
  • SimpleAI Python implementation of many of the artificial intelligence algorithms described on the book “Artificial Intelligence, a Modern Approach”. It focuses on providing an easy to use, well documented and tested library.
  • astroML – Machine Learning and Data Mining for Astronomy.
  • graphlab-create – A library with various machine learning models (regression, clustering, recommender systems, graph analytics, etc.) implemented on top of a disk-backed DataFrame.
  • BigML – A library that contacts external servers.
  • pattern – Web mining module for Python.
  • NuPIC – Numenta Platform for Intelligent Computing.
  • Pylearn2 – A Machine Learning library based on Theano.
  • keras – Modular neural network library based on Theano.
  • Lasagne – Lightweight library to build and train neural networks in Theano.
  • hebel – GPU-Accelerated Deep Learning Library in Python.
  • Chainer – Flexible neural network framework
  • gensim – Topic Modelling for Humans.
  • topik – Topic modelling toolkit
  • PyBrain – Another Python Machine Learning Library.
  • Brainstorm – Fast, flexible and fun neural networks. This is the successor of PyBrain.
  • Crab – A flexible, fast recommender engine.
  • python-recsys – A Python library for implementing a Recommender System.
  • thinking bayes – Book on Bayesian Analysis
  • Restricted Boltzmann Machines -Restricted Boltzmann Machines in Python. [DEEP LEARNING]
  • Bolt – Bolt Online Learning Toolbox
  • CoverTree – Python implementation of cover trees, near-drop-in replacement for scipy.spatial.kdtree
  • nilearn – Machine learning for NeuroImaging in Python
  • imbalanced-learn – Python module to perform under sampling and over sampling with various techniques.
  • Shogun – The Shogun Machine Learning Toolbox
  • Pyevolve – Genetic algorithm framework.
  • Caffe – A deep learning framework developed with cleanliness, readability, and speed in mind.
  • breze – Theano based library for deep and recurrent neural networks
  • pyhsmm – library for approximate unsupervised inference in Bayesian Hidden Markov Models (HMMs) and explicit-duration Hidden semi-Markov Models (HSMMs), focusing on the Bayesian Nonparametric extensions, the HDP-HMM and HDP-HSMM, mostly with weak-limit approximations.
  • mrjob – A library to let Python program run on Hadoop.
  • SKLL – A wrapper around scikit-learn that makes it simpler to conduct experiments.
  • neurolabhttps://github.com/zueve/neurolab
  • Spearmint – Spearmint is a package to perform Bayesian optimization according to the algorithms outlined in the paper: Practical Bayesian Optimization of Machine Learning Algorithms. Jasper Snoek, Hugo Larochelle and Ryan P. Adams. Advances in Neural Information Processing Systems, 2012.
  • Pebl – Python Environment for Bayesian Learning
  • Theano – Optimizing GPU-meta-programming code generating array oriented optimizing math compiler in Python
  • TensorFlow – Open source software library for numerical computation using data flow graphs
  • yahmm – Hidden Markov Models for Python, implemented in Cython for speed and efficiency.
  • python-timbl – A Python extension module wrapping the full TiMBL C++ programming interface. Timbl is an elaborate k-Nearest Neighbours machine learning toolkit.
  • deap – Evolutionary algorithm framework.
  • pydeep – Deep Learning In Python
  • mlxtend – A library consisting of useful tools for data science and machine learning tasks.
  • neon – Nervana’s high-performance Python-based Deep Learning framework [DEEP LEARNING]
  • Optunity – A library dedicated to automated hyperparameter optimization with a simple, lightweight API to facilitate drop-in replacement of grid search.
  • Neural Networks and Deep Learning – Code samples for my book “Neural Networks and Deep Learning” [DEEP LEARNING]
  • Annoy – Approximate nearest neighbours implementation
  • skflow – Simplified interface for TensorFlow, mimicking Scikit Learn.
  • TPOT – Tool that automatically creates and optimizes machine learning pipelines using genetic programming. Consider it your personal data science assistant, automating a tedious part of machine learning.
  • pgmpy A python library for working with Probabilistic Graphical Models.
  • DIGITS – The Deep Learning GPU Training System (DIGITS) is a web application for training deep learning models.
  • Orange – Open source data visualization and data analysis for novices and experts.
  • MXNet – Lightweight, Portable, Flexible Distributed/Mobile Deep Learning with Dynamic, Mutation-aware Dataflow Dep Scheduler; for Python, R, Julia, Go, Javascript and more.
  • milk – Machine learning toolkit focused on supervised classification.
  • TFLearn – Deep learning library featuring a higher-level API for TensorFlow.
  • REP – an IPython-based environment for conducting data-driven research in a consistent and reproducible way. REP is not trying to substitute scikit-learn, but extends it and provides better user experience.

Data Analysis / Data Visualization

  • SciPy – A Python-based ecosystem of open-source software for mathematics, science, and engineering.
  • NumPy – A fundamental package for scientific computing with Python.
  • Numba – Python JIT (just in time) complier to LLVM aimed at scientific Python by the developers of Cython and NumPy.
  • NetworkX – A high-productivity software for complex networks.
  • igraph – binding to igraph library – General purpose graph library
  • Pandas – A library providing high-performance, easy-to-use data structures and data analysis tools.
  • Open Mining – Business Intelligence (BI) in Python (Pandas web interface)
  • PyMC – Markov Chain Monte Carlo sampling toolkit.
  • zipline – A Pythonic algorithmic trading library.
  • PyDy – Short for Python Dynamics, used to assist with workflow in the modeling of dynamic motion based around NumPy, SciPy, IPython, and matplotlib.
  • SymPy – A Python library for symbolic mathematics.
  • statsmodels – Statistical modeling and econometrics in Python.
  • astropy – A community Python library for Astronomy.
  • matplotlib – A Python 2D plotting library.
  • bokeh – Interactive Web Plotting for Python.
  • plotly – Collaborative web plotting for Python and matplotlib.
  • vincent – A Python to Vega translator.
  • d3py – A plotting library for Python, based on D3.js.
  • ggplot – Same API as ggplot2 for R.
  • ggfortify – Unified interface to ggplot2 popular R packages.
  • Kartograph.py – Rendering beautiful SVG maps in Python.
  • pygal – A Python SVG Charts Creator.
  • PyQtGraph – A pure-python graphics and GUI library built on PyQt4 / PySide and NumPy.
  • pycascading
  • Petrel – Tools for writing, submitting, debugging, and monitoring Storm topologies in pure Python.
  • Blaze – NumPy and Pandas interface to Big Data.
  • emcee – The Python ensemble sampling toolkit for affine-invariant MCMC.
  • windML – A Python Framework for Wind Energy Analysis and Prediction
  • vispy – GPU-based high-performance interactive OpenGL 2D/3D data visualization library
  • cerebro2 A web-based visualization and debugging platform for NuPIC.
  • NuPIC Studio An all-in-one NuPIC Hierarchical Temporal Memory visualization and debugging super-tool!
  • SparklingPandas Pandas on PySpark (POPS)
  • Seaborn – A python visualization library based on matplotlib
  • bqplot – An API for plotting in Jupyter (IPython)
  • pastalog – Simple, realtime visualization of neural network training performance.
  • caravel – A data exploration platform designed to be visual, intuitive, and interactive.
  • Dora – Tools for exploratory data analysis in Python.
  • Ruffus – Computation Pipeline library for python.
  • SOMPY – Self Organizing Map written in Python (Uses neural networks for data analysis).
  • HDBScan – implementation of the hdbscan algorithm in Python – used for clustering

Misc Scripts / iPython Notebooks / Codebases

Neural networks

  • Neural networks – NeuralTalk is a Python+numpy project for learning Multimodal Recurrent Neural Networks that describe images with sentences.

Kaggle Competition Source Code

Ruby

Natural Language Processing

  • Treat – Text REtrieval and Annotation Toolkit, definitely the most comprehensive toolkit I’ve encountered so far for Ruby
  • Ruby Linguistics – Linguistics is a framework for building linguistic utilities for Ruby objects in any language. It includes a generic language-independent front end, a module for mapping language codes into language names, and a module which contains various English-language utilities.
  • Stemmer – Expose libstemmer_c to Ruby
  • Ruby Wordnet – This library is a Ruby interface to WordNet
  • Raspel – raspell is an interface binding for ruby
  • UEA Stemmer – Ruby port of UEALite Stemmer – a conservative stemmer for search and indexing
  • Twitter-text-rb – A library that does auto linking and extraction of usernames, lists and hashtags in tweets

General-Purpose Machine Learning

Data Analysis / Data Visualization

  • rsruby – Ruby – R bridge
  • data-visualization-ruby – Source code and supporting content for my Ruby Manor presentation on Data Visualisation with Ruby
  • ruby-plot – gnuplot wrapper for ruby, especially for plotting roc curves into svg files
  • plot-rb – A plotting library in Ruby built on top of Vega and D3.
  • scruffy – A beautiful graphing toolkit for Ruby
  • SciRuby
  • Glean – A data management tool for humans
  • Bioruby
  • Arel

Misc

Rust

General-Purpose Machine Learning

  • deeplearn-rs – deeplearn-rs provides simple networks that use matrix multiplication, addition, and ReLU under the MIT license.
  • rustlearn – a machine learning framework featuring logistic regression, support vector machines, decision trees and random forests.
  • rusty-machine – a pure-rust machine learning library.
  • leaf – open source framework for machine intelligence, sharing concepts from TensorFlow and Caffe. Available under the MIT license. [Deprecated]
  • RustNN – RustNN is a feedforward neural network library.

R

General-Purpose Machine Learning

  • ahaz – ahaz: Regularization for semiparametric additive hazards regression
  • arules – arules: Mining Association Rules and Frequent Itemsets
  • bigrf – bigrf: Big Random Forests: Classification and Regression Forests for Large Data Sets
  • bigRR – bigRR: Generalized Ridge Regression (with special advantage for p >> n cases)
  • bmrm – bmrm: Bundle Methods for Regularized Risk Minimization Package
  • Boruta – Boruta: A wrapper algorithm for all-relevant feature selection
  • bst – bst: Gradient Boosting
  • C50 – C50: C5.0 Decision Trees and Rule-Based Models
  • caret – Classification and Regression Training: Unified interface to ~150 ML algorithms in R.
  • caretEnsemble – caretEnsemble: Framework for fitting multiple caret models as well as creating ensembles of such models.
  • Clever Algorithms For Machine Learning
  • CORElearn – CORElearn: Classification, regression, feature evaluation and ordinal evaluation
  • CoxBoost – CoxBoost: Cox models by likelihood based boosting for a single survival endpoint or competing risks
  • Cubist – Cubist: Rule- and Instance-Based Regression Modeling
  • e1071 – e1071: Misc Functions of the Department of Statistics (e1071), TU Wien
  • earth – earth: Multivariate Adaptive Regression Spline Models
  • elasticnet – elasticnet: Elastic-Net for Sparse Estimation and Sparse PCA
  • ElemStatLearn – ElemStatLearn: Data sets, functions and examples from the book: “The Elements of Statistical Learning, Data Mining, Inference, and Prediction” by Trevor Hastie, Robert Tibshirani and Jerome Friedman Prediction” by Trevor Hastie, Robert Tibshirani and Jerome Friedman
  • evtree – evtree: Evolutionary Learning of Globally Optimal Trees
  • forecast – forecast: Timeseries forecasting using ARIMA, ETS, STLM, TBATS, and neural network models
  • forecastHybrid – forecastHybrid: Automatic ensemble and cross validation of ARIMA, ETS, STLM, TBATS, and neural network models from the “forecast” package
  • fpc – fpc: Flexible procedures for clustering
  • frbs – frbs: Fuzzy Rule-based Systems for Classification and Regression Tasks
  • GAMBoost – GAMBoost: Generalized linear and additive models by likelihood based boosting
  • gamboostLSS – gamboostLSS: Boosting Methods for GAMLSS
  • gbm – gbm: Generalized Boosted Regression Models
  • glmnet – glmnet: Lasso and elastic-net regularized generalized linear models
  • glmpath – glmpath: L1 Regularization Path for Generalized Linear Models and Cox Proportional Hazards Model
  • GMMBoost – GMMBoost: Likelihood-based Boosting for Generalized mixed models
  • grplasso – grplasso: Fitting user specified models with Group Lasso penalty
  • grpreg – grpreg: Regularization paths for regression models with grouped covariates
  • h2o – A framework for fast, parallel, and distributed machine learning algorithms at scale — Deeplearning, Random forests, GBM, KMeans, PCA, GLM
  • hda – hda: Heteroscedastic Discriminant Analysis
  • Introduction to Statistical Learning
  • ipred – ipred: Improved Predictors
  • kernlab – kernlab: Kernel-based Machine Learning Lab
  • klaR – klaR: Classification and visualization
  • lars – lars: Least Angle Regression, Lasso and Forward Stagewise
  • lasso2 – lasso2: L1 constrained estimation aka ‘lasso’
  • LiblineaR – LiblineaR: Linear Predictive Models Based On The Liblinear C/C++ Library
  • LogicReg – LogicReg: Logic Regression
  • Machine Learning For Hackers
  • maptree – maptree: Mapping, pruning, and graphing tree models
  • mboost – mboost: Model-Based Boosting
  • medley – medley: Blending regression models, using a greedy stepwise approach
  • mlr – mlr: Machine Learning in R
  • mvpart – mvpart: Multivariate partitioning
  • ncvreg – ncvreg: Regularization paths for SCAD- and MCP-penalized regression models
  • nnet – nnet: Feed-forward Neural Networks and Multinomial Log-Linear Models
  • oblique.tree – oblique.tree: Oblique Trees for Classification Data
  • pamr – pamr: Pam: prediction analysis for microarrays
  • party – party: A Laboratory for Recursive Partytioning
  • partykit – partykit: A Toolkit for Recursive Partytioning
  • penalized – penalized: L1 (lasso and fused lasso) and L2 (ridge) penalized estimation in GLMs and in the Cox model
  • penalizedLDA – penalizedLDA: Penalized classification using Fisher’s linear discriminant
  • penalizedSVM – penalizedSVM: Feature Selection SVM using penalty functions
  • quantregForest – quantregForest: Quantile Regression Forests
  • randomForest – randomForest: Breiman and Cutler’s random forests for classification and regression
  • randomForestSRC – randomForestSRC: Random Forests for Survival, Regression and Classification (RF-SRC)
  • rattle – rattle: Graphical user interface for data mining in R
  • rda – rda: Shrunken Centroids Regularized Discriminant Analysis
  • rdetools – rdetools: Relevant Dimension Estimation (RDE) in Feature Spaces
  • REEMtree – REEMtree: Regression Trees with Random Effects for Longitudinal (Panel) Data
  • relaxo – relaxo: Relaxed Lasso
  • rgenoud – rgenoud: R version of GENetic Optimization Using Derivatives
  • rgp – rgp: R genetic programming framework
  • Rmalschains – Rmalschains: Continuous Optimization using Memetic Algorithms with Local Search Chains (MA-LS-Chains) in R
  • rminer – rminer: Simpler use of data mining methods (e.g. NN and SVM) in classification and regression
  • ROCR – ROCR: Visualizing the performance of scoring classifiers
  • RoughSets – RoughSets: Data Analysis Using Rough Set and Fuzzy Rough Set Theories
  • rpart – rpart: Recursive Partitioning and Regression Trees
  • RPMM – RPMM: Recursively Partitioned Mixture Model
  • RSNNS – RSNNS: Neural Networks in R using the Stuttgart Neural Network Simulator (SNNS)
  • RWeka – RWeka: R/Weka interface
  • RXshrink – RXshrink: Maximum Likelihood Shrinkage via Generalized Ridge or Least Angle Regression
  • sda – sda: Shrinkage Discriminant Analysis and CAT Score Variable Selection
  • SDDA – SDDA: Stepwise Diagonal Discriminant Analysis
  • SuperLearner and subsemble – Multi-algorithm ensemble learning packages.
  • svmpath – svmpath: svmpath: the SVM Path algorithm
  • tgp – tgp: Bayesian treed Gaussian process models
  • tree – tree: Classification and regression trees
  • varSelRF – varSelRF: Variable selection using random forests
  • XGBoost.R – R binding for eXtreme Gradient Boosting (Tree) Library
  • Optunity – A library dedicated to automated hyperparameter optimization with a simple, lightweight API to facilitate drop-in replacement of grid search. Optunity is written in Python but interfaces seamlessly to R.
  • igraph – binding to igraph library – General purpose graph library
  • MXNet – Lightweight, Portable, Flexible Distributed/Mobile Deep Learning with Dynamic, Mutation-aware Dataflow Dep Scheduler; for Python, R, Julia, Go, Javascript and more.

Data Analysis / Data Visualization

  • ggplot2 – A data visualization package based on the grammar of graphics.

SAS

General-Purpose Machine Learning

  • Enterprise Miner – Data mining and machine learning that creates deployable models using a GUI or code.
  • Factory Miner – Automatically creates deployable machine learning models across numerous market or customer segments using a GUI.

Data Analysis / Data Visualization

  • SAS/STAT – For conducting advanced statistical analysis.
  • University Edition – FREE! Includes all SAS packages necessary for data analysis and visualization, and includes online SAS courses.

High Performance Machine Learning

Natural Language Processing

Demos and Scripts

  • ML_Tables – Concise cheat sheets containing machine learning best practices.
  • enlighten-apply – Example code and materials that illustrate applications of SAS machine learning techniques.
  • enlighten-integration – Example code and materials that illustrate techniques for integrating SAS with other analytics technologies in Java, PMML, Python and R.
  • enlighten-deep – Example code and materials that illustrate using neural networks with several hidden layers in SAS.
  • dm-flow – Library of SAS Enterprise Miner process flow diagrams to help you learn by example about specific data mining topics.

Scala

Natural Language Processing

  • ScalaNLP – ScalaNLP is a suite of machine learning and numerical computing libraries.
  • Breeze – Breeze is a numerical processing library for Scala.
  • Chalk – Chalk is a natural language processing library.
  • FACTORIE – FACTORIE is a toolkit for deployable probabilistic modeling, implemented as a software library in Scala. It provides its users with a succinct language for creating relational factor graphs, estimating parameters and performing inference.

Data Analysis / Data Visualization

  • MLlib in Apache Spark – Distributed machine learning library in Spark
  • Scalding – A Scala API for Cascading
  • Summing Bird – Streaming MapReduce with Scalding and Storm
  • Algebird – Abstract Algebra for Scala
  • xerial – Data management utilities for Scala
  • simmer – Reduce your data. A unix filter for algebird-powered aggregation.
  • PredictionIO – PredictionIO, a machine learning server for software developers and data engineers.
  • BIDMat – CPU and GPU-accelerated matrix library intended to support large-scale exploratory data analysis.
  • Wolfe Declarative Machine Learning
  • Flink – Open source platform for distributed stream and batch data processing.
  • Spark Notebook – Interactive and Reactive Data Science using Scala and Spark.

General-Purpose Machine Learning

  • Conjecture – Scalable Machine Learning in Scalding
  • brushfire – Distributed decision tree ensemble learning in Scala
  • ganitha – scalding powered machine learning
  • adam – A genomics processing engine and specialized file format built using Apache Avro, Apache Spark and Parquet. Apache 2 licensed.
  • bioscala – Bioinformatics for the Scala programming language
  • BIDMach – CPU and GPU-accelerated Machine Learning Library.
  • Figaro – a Scala library for constructing probabilistic models.
  • H2O Sparkling Water – H2O and Spark interoperability.
  • FlinkML in Apache Flink – Distributed machine learning library in Flink
  • DynaML – Scala Library/REPL for Machine Learning Research
  • Saul – Flexible Declarative Learning-Based Programming.

Swift

General-Purpose Machine Learning

  • Swift AI – Highly optimized artificial intelligence and machine learning library written in Swift.
  • BrainCore – The iOS and OS X neural network framework
  • swix – A bare bones library that includes a general matrix language and wraps some OpenCV for iOS development.
  • DeepLearningKit an Open Source Deep Learning Framework for Apple’s iOS, OS X and tvOS. It currently allows using deep convolutional neural network models trained in Caffe on Apple operating systems.
  • AIToolbox – A toolbox framework of AI modules written in Swift: Graphs/Trees, Linear Regression, Support Vector Machines, Neural Networks, PCA, KMeans, Genetic Algorithms, MDP, Mixture of Gaussians.
  • MLKit – A simple Machine Learning Framework written in Swift. Currently features Simple Linear Regression, Polynomial Regression, and Ridge Regression.

TensorFlow

General-Purpose Machine Learning

Credits

表观遗传学药物和免疫治疗方案联合应用,前景值得期待

在2010年,六位晚期肺癌的患者收到了坏消息。临床前小鼠实验中观察到的阿扎胞苷与实验阶段药物恩替诺特联合使用(二者均为表观遗传学修饰药物)对非小细胞肺癌的治疗作用在六位患者体内没有发挥,肿瘤没有任何减小的迹象。为了尽可能延长这些患者的生存期,约翰霍普金斯Sidney Kimmel癌症研究中心的肿瘤学家Julie Brahmer在这六位患者的治疗方案中加入了一种免疫治疗药物——纳武单抗。在其他研究中,以纳武单抗治疗的非小细胞肺癌患者中有接近半数的生存期超过六个月,并且无肿瘤进展。Brahmer及其同事希望她们的六位患者接受nivolumab治疗后,能有三位至少生存六个月,或可能能够生存一年以上。而最终的结果是,五位患者生存期超过了最初的六个月;四年过去了,仍有两位患者生存。其中第三位死亡的患者死于之前治疗的并发症,死亡时并没有明显的肿瘤复发。由于该试验的样本量小,结果的成功不能排除偶然因素。但是能够改变表观基因组的药物阿扎胞苷和恩替诺特,确实可能会促使患者的免疫系统对免疫检查点抑制剂产生应答。这项试验第一次表明了这些药物的配伍能够显著改善患者的临床结果。联合使用不同的治疗药物对抗肿瘤的思路并不新颖。目前科研人员也正尝试将免疫治疗药物,如免疫检查点抑制剂与化疗或放疗等其他治疗手段联合使用。在2015年10月,美国FDA第一次批准了免疫治疗药物的联合治疗方案,即纳武单抗和伊匹单抗联合治疗黑色素瘤。这种联合方案相比单用要更为有效。而表观遗传学药物阿扎胞苷和恩替诺特与免疫治疗药物合用可能会更为有效,在未来几年中会有多项试验来验证这一想法——美国至少有七项,而在加拿大和欧洲有更多项试验正在监测免疫检查点抑制剂和表观遗传学药物联合治疗对抗实体瘤的效果。2015年10月,意大利锡耶纳开展了一项黑色素瘤患者试验。而在约翰霍普金斯,Stephen Baylin团队开始在卵巢癌患者中开展这方面的研究。而前述霍普金斯胸科肿瘤主任Brahmer的研究已进入2期阶段,目前正在Baltimore和加州大学洛杉矶分校招募NSCLC患者。在玛格丽特公主医院安大略肿瘤研究所,Daniel de Carvalho等人准备在多种肿瘤中试验免疫检查点抑制剂和表观遗传学药物合用的效果。表观遗传学药物简述对表观遗传学治疗和免疫治疗联合应用感兴趣的研究人员目前主要研究两大类表观遗传学药物。第一类是组蛋白去乙酰化酶(HDAC)抑制剂,这类药物作用于包裹DNA的组蛋白上,通过控制DNA缠绕于组蛋白的松紧程度来发挥作用。组蛋白去乙酰化酶通过组蛋白的去乙酰化(去除乙酰基),使DNA更紧地缠绕在组蛋白上,从而导致这些DNA不易被基因转录因子接触。结果导致与细胞分化、细胞周期阻滞、肿瘤免疫、受损细胞凋亡等有关的蛋白的表达受到抑制。这些因素都会促使癌症的发展。组蛋白去乙酰化酶抑制剂能有选择性地恢复这些癌症抑制因子和其它抑癌基因的表达。目前FDA已经批准了三个HDAC抑制剂,均用于治疗罕见T细胞淋巴瘤。在约翰霍普金斯试验中加入的恩替诺特就属于这类药物。第二类表观遗传学药物是DNA甲基化抑制剂(DNMTi)。这类药物能够防止肿瘤细胞基因上被加入甲基。DNA甲基化能够沉默延缓细胞分裂的基因,因此可促使细胞增殖不受控制。阿扎胞苷和地西他滨是FDA已经批准的用于骨髓异常增生综合征的两种DNA甲基化抑制剂。虽然表观遗传学药物已经被用于血液肿瘤治疗,但是人们直到最近才开始关注这些药物对免疫系统的影响。一个早期的线索来自2014年的一项研究,荷有乳腺肿瘤或结肠癌细胞的小鼠对免疫检查点抑制剂没有明显应答,但是向免疫治疗方案中加入两种表观遗传学药物之后,80%以上的被治疗小鼠的生存期超过60天,期间没有明显的肿瘤发生,而未治疗小鼠体内则发生明显转移。联合应用的灵感2015年8月份发表的两项研究发现,低剂量阿扎胞苷能够激活细胞警告免疫系统防御病毒入侵的程序。一篇论文结果显示,去甲基化药物如阿扎胞苷能够刺激包括肿瘤细胞在内的人体所有细胞中良性逆转录病毒的表达,这使得肿瘤细胞更加具有免疫原性,吸引免疫细胞的能力更强。在另一项研究中,De Carvalho等人发现了相同的病毒反应。表观遗传学药物仅能较弱地刺激免疫系统,在没有免疫检查点抑制剂时,肿瘤依然能够抑制免疫应答。也就是说,虽然表观遗传学治疗已经使肿瘤细胞更具免疫原性,但是加入免疫检查点抑制剂去除免疫系统的“刹车”后,T细胞才能够攻击肿瘤细胞。一些研究人员认为DNA甲基化抑制剂(DNMTis)比组蛋白去乙酰化酶(HDAC)抑制剂更适合与免疫治疗配伍。一些动物研究结果显示,HDAC抑制剂脱靶效应明显,因此副作用往往较重,包括疲劳和恶心等。甲基化抑制剂在血液肿瘤治疗的临床数据也更优于HDAC抑制剂,在一项2期骨髓异常增生综合征和急性髓系白血病试验中,对HDAC抑制剂产生治疗应答的患者百分率很低,但是对DNMT抑制剂产生应答的患者却达到40%。Baylin团队认为两种表观遗传学药物合用效果会由于一种药物单用,2011年研究以及其未发表的小鼠数据都支持这一观点。因此,他的团队计划开展NSCLC临床试验来比较不同药物合用的疗效:一种免疫抑制剂与一种DNMT抑制剂,或一种HDAC抑制剂,或与后两者合用。而Brahmer坚持采用“阿扎胞苷-恩替诺特-纳武单抗”这一在之前的研究中已显示良好效果的合用策略。联合用药方案的实施如果以上描述的合用策略被证实有效,那么临床医生可能不得不要重新定义治疗成功的含义。传统的治疗起效的标准——快速的肿瘤缩减,可能不再适合评价免疫治疗联合方案的效果。2011年,Baylin在约翰霍普金斯主持开展的临床研究显示,仅接受表观遗传学药物治疗的患者基本没有获益,仅有少部分患者的肿瘤可见明显缩减。另外,免疫治疗使肿瘤发生缩减之前,有时还能够观察到肿瘤的增大现象。其中六位患者——即本文开篇所述的Brahmer的六位患者开始接受免疫检查点抑制剂进行治疗,治疗效果如本文开篇所述令人振奋。这时研究人员开始提出表观遗传学药物能够使肿瘤细胞对免疫治疗更为敏感的设想。因此,相比肿瘤缩减,六位患者试验得出的“六个月无进展生存”可能更适合作为治疗成功的标准。当前,与免疫治疗联合应用的治疗方案还包括化疗、放疗及靶向治疗,这些药物作用方式与表观遗传学药物不同,但是也可引发免疫应答。当肿瘤细胞被放化疗杀死后,免疫细胞也被募集而来,引发类似自身免疫疾病中的炎症反应。表观遗传学药物相比放化疗更具有选择性,因为这些药物识别的表观遗传学标志物大多出现在肿瘤细胞中。这种选择性可能会减少自身免疫疾病样的副作用。按照Baylin的说法,具体的给药方案为,给予表观遗传学药物后间隔较短时间便给予免疫检查点抑制剂,表观遗传学治疗先使肿瘤细胞更具免疫原性,然后免疫检查点抑制剂去除免疫系统的“刹车”,最后免疫系统攻击肿瘤细胞。以上所述的所有联合治疗成功与否还需等待患者的临床结果,在未来一年中结果会逐渐明朗。Baylin说,他感到乐观的同时也感觉害怕,“在你没有看见真正的结果之前,一切都只是猜想。”希望在更多的患者中重复出在最初六位肺癌患者中取得的成功结果。

参考文献:Karen Weintraub. Take two: Combining immunotherapy with epigenetic drugs to tackle cancer. Nature Med, Jan 2016, Volume 22, Number 1.

癌症免疫疗法“最新综述”TOP8

近几年,癌症免疫疗法进入飞速发展阶段,然而尽管这一突破性的技术给科学界带来了很大惊喜,也吸引了一大批制药企业加入这一领域。

近几年,癌症免疫疗法进入飞速发展阶段,然而尽管这一突破性的技术给科学界带来了很大的惊喜,也吸引了一大批制药企业加入这一领域,但包括检查点抑制剂、 过继细胞疗法、癌症疫苗在内的各项技术仍存在很多需要克服的障碍。近期,Cell、Nature等杂志发表了多篇癌症免疫综述,共同探讨了推动免疫治疗发 展的方法。

Safety profiles of anti-CTLA-4 and anti-PD-1 antibodies alone and in combination

5月4日,在线发表在Nature Reviews Clinical Oncology上的这篇综述指出,PD-1和CTLA-4抗体为晚期黑色素瘤患者带来了前所未有的希望,并且在其它癌症类型中也显示出了巨大的潜力。然 而,这些新型免疫疗法也带来了一些独特的不良反应,PD-1和CTLA-4抑制剂的副作用通常出现在皮肤、胃肠道、肝脏和内分泌系统,包括瘙痒、皮疹、恶 心、腹泻和甲状腺疾病。这一综述中,作者们概括了靶向PD-1和CTLA-4的检查点抑制剂的不良事件,旨在提供一些通用准则。[文献]

Adapting Cancer Immunotherapy Models for the Real World

4月19日,在线发表于Cell旗下Trends in Immunology上的这篇综述文章中,作者们讨论了患者差异如何影响癌症免疫疗法的疗效和毒性,以及如何根据不同的宿主环境在动物研究中更好的建模。

文章指出近期的小鼠研究表明,年龄、肥胖以及微生物群对癌症的天然免疫力以及响应免疫疗法的能力有深远的影响。尽管这一研究领域正处于起步阶段,但这些结 果足以支撑该研究方向,即人类癌症免疫疗法如何利用小鼠更好的建模。作者们认为,只有系统测试了各种不同类型的小鼠(年轻/年老、胖/瘦、携带不同微生物 群)才能真正揭开人类癌症免疫疗法的复杂性。[详细]

Mechanism-driven biomarkers to guide immune checkpoint blockade in cancer therapy

4月15日,在线发表于Nature Reviews Cancer上的这篇综述指出,尽管近几年多个抑制CTLA-4和PD-1的抗体获批,且一些其它的免疫检查点正在临床验证中,但通过阻断检查点通路的抗癌药的优化使用依然存在问题。

文章强调,目前,除CTLA4和PD1抗体已经获得批准外,一些其它检查点受体和配体靶向的临床试验也在不断增加,包括LAG3、TIM3、 B7H3(CD276)、CD39、CD73以及腺苷A2a受体。大多数这些免疫检查点的开发结合了PD-1通路抑制抗体。这其中的一些检查点与PD- L1共表达,为这类双重阻断疗法提供了依据。然而,由于相关的临床试验还处于早期阶段,还没有已经经验证的生物标志物可以用于预测哪些病人可以更加受益于 这种双重抑制疗法。[详细]

The future of cancer treatment: immunomodulation, CARs and combination immunotherapy

3月15日,在线发表于Nature Reviews Clinical Oncology上的这篇综述回顾了癌症免疫治疗的最新进展,讨论了这一疗法在未来癌症治疗中的角色,概述了组合方案潜在的治疗相关性。该综述主要介绍了 三方面的内容:单抗的免疫调节、CAR-T疗法以及联合治疗。文章详细总结了检查点阻断疗法以及CAR-T疗法涉及的多种新靶点。[详细]

Inducing stable reversion to achieve cancer control Vaccines for established cancer: overcoming the challenges posed by immune evasion

3月11日,在线发表于Nature Reviews Cancer上这篇综述中,作者们讨论了克服由肿瘤细胞内在因子和肿瘤微环境控制的免疫逃逸相关的过程,总结了如何通过改善疫苗设计、联合疫苗与标准化疗 等方式使免疫治疗效益最大化。这一综述分析了一些成功免疫疗法的共性,主要可分为以下三个因素:

1. 肿瘤相关抗原的选择,这取决于癌症的类型。在病毒诱导的癌症类型中,抗原也应该来自病毒;在其它起源的癌症类型中,应选择neo-antigens(源于肿瘤细胞的突变)。

2. 创建一个能够诱导产生正确免疫细胞的平台,包括产生比例平衡的效应细胞(CD4/CD8+)以及记忆T细胞。

3. 成功的免疫治疗是靶向肿瘤细胞的免疫抑制策略,这可以通过特定的联合治疗实现,比如联合化疗、检查点抑制剂等。[详细]

Development of immuno-oncology drugs—from CTLA4 to PD1 to the next generations

3月11日,在线发表于Nature Reviews Drug Discovery上的这篇综述回顾了癌症免疫疗法近年来的发展历史,包括成功的因素;概述了新药研发的注意事项,总结了自2011年发展起来的三代免疫 疗法,说明了这些新一代免疫疗法将带来的新机会。[详细]

The Basis of Oncoimmunology

3月10日,发表于Cell上的这一综述中,作者讨论了癌症免疫响应的组成、基于TH2的抗癌疗法、T细胞免疫靶向治疗、微生物在调节系统癌症风险和响应治疗中的作用以及肿瘤免疫治疗模式等内容。[详细]

Anti-CD73 in Cancer Immunotherapy: Awakening New Opportunities

2月4日,在线发表于Cell旗下Trends in Cancer上的这一综述讨论了CD73与肿瘤发生、发展以及扩散之间的关系,强调了这一分子作为药物靶标的潜在价值,并表示CD73有望成为个性化癌症治疗中的新生物标志物。[详细]

四项最新进展

近期,除了不断有科学家在这些顶级学术期刊上发表综述文章外,癌症免疫疗法领域也取得了多项新进展,其中“借助他人免疫细胞对抗癌症”的研究引起了广泛的 关注。5月19日,发表在Science上的一项研究中,科学家们发现即使患者自身的免疫细胞不能识别和对抗他们的肿瘤,但其他人的免疫细胞也许可行。

研究人员测试了是否相同的neoantigens能够被来源健康捐赠者的T细胞识别。让人惊讶的是,这些捐赠者来源的T细胞能够检测出非常多患者T细胞不 能识别的neoantigens。领导该研究的Ton Schumachervia表示,依据这些捐赠T细胞使用的受体可以工程改造患者自身的T细胞,使它们能够检测到癌细胞。[详细]

5月23日,发表在Nature上的一项研究中,科学家们发现了一些癌细胞中的遗传变异能够导致PD-L1蛋白表达增强。研究小组描述了他们对成人T细胞白血病/淋巴瘤病例的测序结果,发现这些变异有望被用作癌症患者的生物标志物。[详细]

3月24日,发表在Cell杂志上的一项研究中,研究人员分析了治疗前黑色素瘤活组织样本的mutanomes和转录组,以期鉴定出影响患者对PD-1疗 法敏感性或抵抗性的因素。结果发现,整体高突变负荷与改善生存相关,响应PD-1疗法的患者拥有丰富的BRCA2突变。[详细]

3月3日,发表在Science杂志上的一项中,由英国伦敦大学学院领导的科学家小组找到了能够保证免疫疗法实现精准治疗的新依据。他们鉴定出的neoantigens携带了癌症发展最早期的突变,且在所有肿瘤细胞中都有表达,而不是某个子集。

Cell综述:为何“癌症免疫疗法”只对部分人有效?

Mutations Associated with Acquired Resistance to PD-1 Blockade in Melanoma

上周,笔者分享了Nature Reviews Clinical Oncology上发表的一篇CAR-T疗法相关综述,文章总结了该疗法在研的血液学肿瘤和实体肿瘤相关靶点,讨论了改善这一疗法的6大途径,并强调了在实现最大抗肿瘤效力的情况下平衡安全性需考虑的3方面因素。今天,为大家介绍一篇Cell子刊上发表的免疫检查点阻断疗法相关的最新综述。

近 几年,免疫检查点抑制剂在多种癌症类型中取得了惊人的治疗效果,现已成为最前沿的癌症免疫疗法之一。6月21日,Cell旗下Immunity杂志(最新 影响因子24.082)发表了题为“Resistance Mechanisms to Immune-Checkpoint Blockade in Cancer: Tumor-Intrinsic and -Extrinsic Factors”的综述文章。

作者们总结了癌细胞自主因 素、肿瘤微环境因素以及宿主相关因素如何在癌症治疗时影响免疫检查点阻断疗法,使其呈现出多样化的响应。此外,文章还讨论了免疫系统与宿主微生物群之间的 互相关系能够决定癌症治疗响应的新证据。作者们提出了一个概念,即在使用免疫检查点阻断疗法之前或治疗期间调节肠道微生物可优化治疗效果。

免疫检查点疗法进展回顾

过 去十年中,癌症免疫疗法前所未有的崛起和成功彻底改变了多种恶性肿瘤的临床管理。其中,免疫检查点阻断剂(Immune-Checkpoint Blockers,ICBs)是免疫疗法中最前沿的技术之一。这类疗法在多种组织学肿瘤类型中具有广泛的生物活性,且反应持久。根据目前的临床结果,免疫 检查点阻断疗法中最成功的要属靶向CTLA-4和PD-1/PD-L1的药物。

引言中,作者们回顾了CTLA-4、 PD-1/PD-L1抗体的作用机理和发展历程。截止到目前(作者统计时),FDA批准的ICBs共4个,分别为,1)ipilimumab:CTLA- 4单抗,用于治疗不可切除或转移性黑色素瘤;2)pembrolizumab:PD-1单抗,用于治疗不可切除的转移性黑色素瘤以及PD-L1阳性晚期转 移性非小细胞肺癌;3)nivolumab:PD-1单抗,用于治疗不可切除或转移性黑色素瘤、接受含铂化疗治疗期间或治疗后病情进展的晚期转移性非小细 胞肺癌以及晚期(转移性)肾细胞癌;4)atezolizumab:PD-L1抗体,用于治疗对含铂化疗不响应的局部晚期或转移性尿路上皮癌。

然 而,尽管这类抗体对癌症治疗起到了很强的改善作用,但不可忽视的是,大多数患者未能响应ICBs,甚至因为免疫相关不良事件(immune- related adverse events,irAEs)的发生必须停止治疗。对于ICBs来说,虽然一些数据证实了它前所未有的实力,但截止目前的临床数据显 示,ipilimumab治疗的患者响应率约为15%;而靶向PD-1/PD-L1的药物治疗的患者响应率很少超过40%,且其中有大量的部分应答。因 此,有2个重要的问题需要解决:1)为什么患者响应ICBs会产生一定程度的异质性;2)ICBs如何才能将覆盖范围扩展到多数癌症患者群体中?

为何产生耐药性?

Major Factors Contributing to Primary Resistance to ICB Therapy(图1)

作 者们认为,从某种程度上来说,通过进一步深入理解和研究肿瘤微环境(tumor microenvironment,TME)中的免疫调节机制,这些问题的答案最终会浮出水面。TME中有很多因素能够抑制ICBs的治疗活性,如 Treg细胞、髓系来源抑制细胞(MDSCs)以及IDO(indole 2,3-dioxygenase)的活性;同时,一些肿瘤细胞自身因素也会影响ICBs的疗效,包括突变负荷、致瘤信号通路、PD-L1的表达以及 MHC-I类分子表达下调等。

事实上,肿瘤内在的因素并不是影响免疫疗法结果的唯一因素(图1)。一些新证据表明,癌症免疫疗法还受到影响 免疫系统功能的宿主相关和环境因素的影响。在这一综述中,作者们汇总了这些较少被考虑到的、有可能决定免疫检查点阻断疗法成败的肿瘤外在(tumor- extrinsic)因素。同时,鉴于一些新的发现,他们特别注意到了肠道微生物群的免疫调节潜能。

正文中,作者们 以大量的文献分别从TME(第一部分)、癌细胞自发机制(第二部分)两方面解释了免疫检测阻断疗法的耐药性。具体来说,TME部分相关的内容包括免疫调节 通路、Th1和 Tc1趋化因子分泌的表观遗传沉默以及I型干扰素信号的重要性;癌细胞自发机制部分介绍了致癌信号、突变状态、炎症和代谢线索等内容。

Immune-Checkpoint Blockade Mobilizes the Gut Microbiota to Promote Anti-tumor Immune Responses(图2)

第 三部分内容介绍了导致不良免疫疗法响应的宿主相关因素,如年龄、HLA分型、遗传多态性、饮食和新陈代谢以及慢性感染背景。第四部分内容中,作者们强调了 肠道微生物组对成功的癌症免疫疗法的重要性,分析了肠道微生物组对ICB免疫疗法的影响,并称它们可能与免疫疗法引发的胃肠道毒性相关,未来有希望能够通 过进一步研究解决这一问题(图2)。

Mobilizing the Gut Microbiota to Circumvent Primary Resistance to ICB in Patients(图3)

第 四部分内容(图3)汇总了改善免疫检查点阻断疗法覆盖范围的新途径。一方面,作者们介绍称,可以利用现有的治疗手段帮助扫除免疫疗法的障碍;另一方面,他 们再次强调可以通过操纵微生物组扩大ICB的治疗范围;此外,文章还提出了一种称作“Oncomicrobiotics”概念。

结论

大 量肿瘤或宿主相关的因素通过不同的方式组合决定了ICB抗癌疗法临床响应的异质性。近几年,科学家们通过将ICBs结合各种辅助疗法,改善其持久性、疗效 等属性,以期降低这种异质性。其中,有很多途径成功了,尽管通常毒性也会随之增加。值得注意的是,操纵肠道微生物相关的成果极为引人注目。作者们还认为, 饮食、益生菌或选择性抗生素的管理以及特殊菌株(oncomicrobiotics)或其产物的补充应该被考虑为一种组合策略,用以支持肠道免疫力,刺激 有效的抗癌免疫监视。

 

Mutation per Mb DNA was calculated by dividing the total number of somatic SNVs and indels overlapping with the coding regions and essential splice sites by the total number of coding bases suf ciently covered (≥3× in tumor and ≥10× in matched normal samples) by sequencing data.

抗PD-1免疫治疗获得性耐药新发现

近年来,促进免疫系统防御肿瘤细胞的抗癌药物,如百时美施贵宝的Opdivo和默沙东的Keytruda,给全世界的癌症患者带来了莫大希望,它们能介导 持久的病情缓解,甚至可能攻克某些癌症。与此同时,一块乌云也悬挂在肿瘤免疫疗法领域的上空:研究人员发现,某些患者可能在首次用药时有良好反应,但他们 的癌症在一段疗效期后会复发。
近日,美国加州大学洛杉矶分校(University of California, Los Angles)的研究人员首次探明了晚期黑色素瘤是如何对免疫疗法产生耐药性的。这项主要由美国国家癌症研究所资助的研究,以论文形式发表在了最新一期《新英格兰医学杂志》上。

Antoni Ribas博士

我 们对本研究的领导者并不陌生,他就是享誉盛名的加州大学洛杉矶分校肿瘤免疫学项目主任Antoni Ribas博士,他曾是抗PD-1药物Keytruda(pembrolizumab)大型1期临床阶段试验的主要研究者。作为一种免疫检查点抑制 剂,Keytruda结合T细胞上的PD-1分子,激活被压制的免疫反应,增强机体本身的免疫功能来防御肿瘤细胞。他对这项题为“Mutations Associated with Acquired Resistance to PD-1 Blockade in Melanoma”的研究评论说:“这将有助于我们更好地设计下一代的肿瘤免疫治疗。”

PD-1信号通路抑制免疫反应

据估计,40%左右的晚期黑色素瘤(最致命的皮肤癌形式)患者会最初对免疫疗法有反应,但是四分之一的病人会在三年治疗期间内复发。

复发的晚期黑色素瘤细胞

加 州大学洛杉矶分校的研究人员搜集了复发病人的黑色素瘤活体组织切片,并且分析比较了Keytruda治疗前后肿瘤细胞的全基因组序列。其中,有一个病人的 肿瘤细胞已经失去了一个称为B2M的基因,它编码“抗原提呈蛋白β2-微球蛋白”(antigen-presentingproteinbeta-2- microglobulin),于是改变了免疫系统识别癌细胞的方式。另外两个患者肿瘤发生了JAK1和JAK2基因的功能缺陷,限制了免疫系统杀死癌细 胞的能力。

全基因组分析肿瘤细胞,筛选突变基因

虽然这些患者使用Keytruda治疗,Antoni Ribas博士认为结果可以推广到所有的PD-1相关疗法。他还表示,鉴定耐药机制可能有助于解释为什么有些患者不对免疫疗法有任何响应。
“如果我们深入了解耐药机理,我们也许能够设计更好的治疗方案。还需要做很多工作来研发靶向这些突变的药物。”
的确,精准医疗的精髓就在于深入了解每个病人的具体遗传背景,设计最个体化的医疗方案来最快、最准确、最安全有效的预防治愈病患。我们非常高兴看到日新月异的科技发展,推动全世界人民更美好的生活质量。
参考资料:
[1] UCLA study finds why some cancers stop responding to immunotherapy
[2] Mutations Associated with Acquired Resistance to PD-1 Blockade in Melanoma- the New England Journal of Medicine
[3] UCLA官方网站

近日,美国加州大学洛杉矶分校的AntoniRibas教授团队首次探明了晚期黑色素瘤接受抗PD-1免疫治疗后出现获得性耐药的新机制。该研究发表在了 最新一期《The New England Journal of Medicine》上。该研究通过搜集黑色素瘤病人治疗前及复发后组织标本,采用全外显子测序技术,分析比较了pembrolizumab(anti- PD-1)治疗前及复发耐药后肿瘤细胞的全基因组序列。结果分析发现两个患者肿瘤发生了JAK1和JAK2基因的功能缺陷,第三个患者则出现B2M基因突 变。研究进一步表明前者主要参与免疫系统对肿瘤细胞的杀伤作用,而后者则主要通过影响抗原提呈从而参与免疫系统对肿瘤细胞的识别。这一新的发现对于未来攻 克免疫治疗耐药这一领域将提供重要的理论依据,同时也将有助于我们更好地设计下一代的肿瘤免疫治疗药物。

该研究从15例接受pembrolizumab治疗并出现局部缓解及复发进展的黑色素瘤患者中筛选出4例患者,对其治疗前及复发耐药后肿瘤标本进行全外显 子测序。作者首先对4例筛选患者不同治疗时段(baseline,activeresponse 和relapse)的肿瘤标本进行相关免疫分子检测。结果提示基线状态下PD-L1与CD8+ T细胞主要共定位存在肿瘤外侵边缘,当肿瘤获得有效控制时则主要表现为CD8+ T细胞明显增多,而当肿瘤出现复发进展时,肿瘤细胞PD-L1表达阴性,肿瘤边缘的巨噬细胞及基质细胞PD-L1表达阳性。不同治疗时间点PD-L1及 CD8表达的动态变化也反应了机体免疫微环境发生了相应改变。          作者通过分析肿瘤标本测序结果发现1号患者及2号患者复发后肿瘤标本分别存在JAK1及JAK2基因突变,且都同时伴有相同染色体杂合性缺失 (Lossof heterozygosity,LOH)。作者进一步从体外细胞实验探讨了JAK1/JAK2基因功能缺失对机体免疫抗肿瘤反应的影响。结果分析复发后黑 色素瘤细胞株M464(JAK2突变)与基线黑色素瘤细胞株M420相比,Interferon(IFN)-γ信号通路(JAK2/STAT/IRF1) 失活及免疫相关分子PD-L1和MHC-I表达下调。
体外增殖实验也发现,相比M420(基线,JAK1/JAK2野生型),使用IFN-γ刺激不能抑制M464(JAK2突变)细胞的增殖能力,而另一株复 发后黑色素瘤细胞株M407(JAK1突变)对IFN-α/β/γ刺激均不敏感。这一结果进一步明确了JAK1/JAK2基因突变直接导致了肿瘤细胞对 IFN的杀伤作用不敏感,从而促使肿瘤细胞对PD-1抑制剂的耐药抵抗。         3号病人复发肿瘤外显子测序表明存在B2M基因的突变,B2M是MHC-I分子的组成部分,在免疫治疗抗原提呈过程中发挥重要作用。以前多项研究也 早已表明B2M基因缺陷是免疫治疗获得性耐药的一个重要机制。且该例患者基线及复发肿瘤MHC-II均表达阴性,提示在MHC-I功能缺陷的状态下 MHC-II分子也不能发挥代偿作用。因此,B2M基因的突变可能是导致该患者出现anti-PD-1耐药的主要原因。        在今年3月份《naturecommunication》杂志上同样发表了一篇关于免疫治疗获得性耐药的文章,该文从另一个角度阐述了免疫治疗获得 性耐药机制——TIM3“旁路激活”。目前关于免疫治疗耐药机制的讨论成为继EGFR-TKI耐药之后又一大热门话题。归纳讨论结果主要集中在以下三部 分:(1)非PD-L1的其他抑制性checkpoints继发性过表达;(2)抗原提呈信号通路异常;(3)T细胞活化及杀伤功能异常。        考虑到本研究纳入的样本量过少,是否可以从结论推至大众还无从说起。期待未来纳入更多此类患者,包括不同瘤种、不同免疫治疗耐药模型以便更深入探讨 免疫治疗的耐药机制,进而开发新一代的针对获得性耐药的免疫治疗药物。

Lancet:抗HER2合用方案可作为结直肠癌的一二线治疗方案?

近期HERACLES和MyPathway basket两项临床试验表明,曲妥珠单抗与拉帕替尼(或帕妥珠单抗)合用抑制HER2在结直肠癌中的治疗效果优于标准化疗方案。目前已经提出是否能够将曲妥珠单抗合用方案作为结直肠癌的一二线治疗方案,相关试验正在进行当中。不仅如此,联合抑制EGFR和HER2-4的用药方案也显示出良好的前景。

晚期结直肠癌预后差,亟待出现更有活性的药物。HER2是结直肠癌中一个较为理想的靶点。在患者来源的移植瘤模型或HER2扩增的转移性结直肠癌模型中获得的临床前数据显示,曲妥珠单抗和拉帕替尼合用有较好的应用前景。在HERACLES这项2期临床试验中,Andrea Sartore-Bianchi等研究人员分析了曲妥珠单抗与拉帕替尼合用在有HER2扩增或过表达的难治性结直肠癌患者中的效果。Sartore-Bianchi等人筛选了914位化疗难治性且KRAS野生型结直肠癌患者,在其中鉴定出46位HER2阳性患者,最终招募27位合格患者进行本项研究。其中,8名患者(30%)出现客观缓解(这是在过往难治性患者中所报道的最高比例),一名患者(4%)达到完全缓解,20名患者(74%)或为完全缓解,或为部分缓解,或达到疾病稳定状态。治疗反应持续时间中位数为9.5个月,无进展生存期中位数为5.2个月,总生存期中位数为11.5个月。这项试验结果发表于在Lancet Oncology上。

尽管这项试验的样本量较小,其结果还是很有突破性的,显示了HER2是结直肠癌治疗的一个理想靶点。这项结果与MyPathway basket试验的数据一致。MyPathway basket试验是在HER2、BRAF、EGFR或hedgehog阳性的难治性结直肠癌患者中进行的。研究人员给予13名患者曲妥珠单抗(靶向HER2)和帕妥珠单抗(抑制HER2-HER3二聚作用和HER2激活)合用,结果与HERACLES试验结果相似。这两项试验结果表明,将曲妥珠单抗与拉帕替尼或帕妥珠单抗合用抑制HER2的治疗效果优于标准化疗方案。尽管样本量较小,这些数据为HER2扩增或过表达的结直肠癌患者提供了一种新的常规治疗标准方案:将anti-HER2(曲妥珠单抗或拉帕替尼)与anti-HER2-3(帕妥珠单抗)或anti-HER1/EGFR(拉帕替尼)合用对于结直肠癌治疗是一个十分有效的策略。

在得出以上令人鼓舞的结论的同时,还存在几个问题需要解释或进一步研究。

首先是联合方案如何选择的问题。临床前数据表明,RAS突变肿瘤中的HER2过表达可能性较小。由于拉帕替尼可抑制HER2和EGFR,而帕妥珠单抗不能抑制EGFR,因此曲妥珠单抗与拉帕替尼合用,相比与帕妥珠单抗合用可能在KRAS突变的患者中有更好的效果。目前来讲,曲妥珠单抗与拉帕替尼合用是较好选择,但是帕妥珠单抗可能是一个毒性较低的备选药物。

第二个问题是anti-HER2治疗是否应该作为一二线治疗方案。西妥昔单抗治疗无效的患者(无论是否有KRAS突变)均对anti-HER2补救疗法敏感,因此曲妥珠单抗的联合用药方案应该放在EGFR抗体之前考虑,甚至可以作为一线治疗方案。例如在MODUL试验中,氟尿嘧啶和曲妥珠单抗及帕妥珠单抗合用被用作一线维持治疗方案。

第三,HERACLES试验显示,曲妥珠单抗合用治疗在HER2免疫组化评分超过3分的肿瘤中更有效。但即便是在HER2表达较低的患者中,曲妥珠单抗合用也会显示出治疗效果。因此, HER2评分2+或3+的患者应该接受靶向HER2治疗

第四个问题是,HERACLES试验的研究人员给出了HER2阳性的诊断标准,应该采用这个标准来判断可能会从曲妥珠单抗联合治疗中受益的患者。

第五,选择anti-HER2治疗方案之前是否还需要重新取材检测HER2。回顾性分析数据和HERACLES试验结果显示,原发肿瘤和转移灶的HER2状态有很好的一致性,因此重新取材可能没有必要。如果需要检测的话,液体活检可能比较适用于这种情况。

最后一个问题是,anti-HER2治疗的后续研究还有哪些?随着Anti-HER2治疗可以作为常规的补救疗法,靶向HER3以及联合靶向EGFR和HER2-4的试验也在进行当中。HER3在结直肠癌中的表达较为常见(大约占75%),使其成为一个受到关注的靶点。NCI-NSABP FC7试验正在研究来那替尼(影响EGFR、HER2和HER4的酪氨酸激酶抑制剂)和西妥昔单抗合用的疗效,初步显示效果良好,且不受RAS突变的影响。临床前和早期临床数据均显示,联合抑制EGFR和HER2-4的方案有良好的前景。

Targeting HER2: precision oncology for colorectal cancer. Lancet Oncol 2016. Published online April 20, 2016 http://dx.doi.org/10.1016/S1470-2045(16)30039-0.

Advanced colorectal cancer still has a poor prognosis and more active drugs are urgently needed. HER2 was investigated as a target in colorectal cancer in two early trials of trastuzumab plus chemotherapy as first, second or third line therapy, which produced interesting but conflicting results.1,2 In The Lancet Oncology, Andrea Sartore-Bianchi and colleagues3 present the results of the HERACLES trial, the first phase 2 trial in patients with refractory colorectal cancer and HER2 amplification or overexpression.