Detecting repeated cancer evolution from multi-region tumor sequencing data.
Detecting repeated cancer evolution from multi-region tumor sequencing data.
复制标题
DOI：
10.1038/s41592-018-0108-x
复制
发表时间：
2018-09
期刊：
影响因子：
48
通讯作者：
Sottoriva A
中科院分区：
文献类型：
--
作者：
Caravagna G;Giarratano Y;Ramazzotti D;Tomlinson I;Graham TA;Sanguinetti G;Sottoriva A
癌症演变是由（表观）基因组畸变的积累驱动的。患者之间和患者内部的基因组变化的重复序列反映了重复的进化，这对于预测癌症进展是有价值的。多区域测序允许推断肿瘤内基因组变化的一些时间顺序。然而，进化过程固有的随机性使得不同的患者看起来非常不同，阻止了重复进化的鲁棒识别。在这里，我们提出了一种基于迁移学习的新型机器学习方法，该方法克服了癌症演变和数据中噪声的随机效应，突出了癌症队列中隐藏的进化模式。当应用于来自肺癌、乳腺癌、肾癌和结直肠癌的多区域测序数据集（来自178名患者的768个样本）时，我们的方法检测到了患者亚组中重复的进化轨迹，这些轨迹在单样本队列（n= 2，935）中重现。我们的方法提供了一种新的方法，可以根据肿瘤的演变对患者进行分类，并对预测癌症演变产生影响。
Cancer evolution is driven by the accumulation of (epi)genomic aberrations. Recurrent sequences of genomic changes, between and within patients, reflect repeated evolution that is valuable for anticipating cancer progression. Multi-region sequencing allows inference of some temporal orderings of genomic changes within a tumour. However, the inherent stochasticity of the evolutionary process makes different patients appear very distinct, preventing the robust identification of repeated evolution. Here we present a novel machine learning method based on Transfer Learning that overcomes the stochastic effects of cancer evolution and noise in the data, highlighting hidden evolutionary patterns in cancer cohorts. When applied to multi-region sequencing datasets from lung, breast, renal and colorectal cancer (768 samples from 178 patients), our method detected repeated evolutionary trajectories in subgroups of patients, which reproduced in single-sample cohorts (n=2,935). Our method provides novel ways to classify patients based on how their tumour evolved, with implications for anticipating cancer evolution.