已知异常 DNA 甲基化 (DNAm) 与癌症的病因有关，包括结直肠癌 (CRC)。过去，开放获取数据的可用性一直是创新方法开发和研究培训的主要驱动力。然而，这一点正日益受到受控访问的影响，尤其是医疗数据，包括癌症 DNAm 数据。为了复兴这一宝贵的传统，我们利用了来自 NCBI GEO 和 ArrayExpress 的 14 项开放获取研究的 1,845 个样本（535 个 CRC 肿瘤、522 个肿瘤附近的正常结肠组织、72 个结直肠腺瘤和来自健康个体的 716 个正常结肠组织）的 DNAm 数据。我们使用十一个表观遗传时钟模型计算了每个样本的表观遗传年龄（EA），并得出相应的表观遗传年龄加速（EAA）。对于 EA，我们观察到大多数第一代和第二代表观遗传时钟反映了肿瘤附近正常组织和健康个体的实际年龄 [例如，Horvath（r = 0.77 和 0.79）、Zhang 弹性网（EN）（r = 0.70）和 0.73)] 与表观遗传有丝分裂时钟 (EpiTOC、HypoClock、MiAge) (r < 0.3) 不同。对于 EAA，我们使用 PhenoAge、Wu 和上述有丝分裂时钟，发现它们在不同组织类型中具有不同的分布，特别是在邻近肿瘤的正常结肠组织和癌性肿瘤之间，以及邻近肿瘤的正常结肠组织和正常结肠组织之间来自健康个体的结肠组织。最后，我们利用这些关联开发了一个使用弹性网络回归（具有套索和岭正则化）的分类器，该分类器根据患者性别和从组织学正常对照（即与肿瘤相邻的正常结肠组织和正常结肠组织）计算出的 EAA 来预测 CRC 诊断来自健康个体）。该分类器表现出良好的诊断潜力，ROC-AUC = 0.886，这表明基于 EAA 的相关数据训练的分类器可以成为临床专业人员支持 CRC 诊断/预后决策的工具。我们的研究还再次强调了开放获取临床数据对于方法开发和年轻科学家培训的重要性。在本研究的时间范围内获得受控访问数据所需的批准是不可能的。版权所有 © 2023 Widayati、Schneider、Panteleeva、Chernysheva、Hrbkova、Beck、Voloshin 和 Chervova。

Aberrant DNA methylation (DNAm) is known to be associated with the aetiology of cancer, including colorectal cancer (CRC). In the past, the availability of open access data has been the main driver of innovative method development and research training. However, this is increasingly being eroded by the move to controlled access, particularly of medical data, including cancer DNAm data. To rejuvenate this valuable tradition, we leveraged DNAm data from 1,845 samples (535 CRC tumours, 522 normal colon tissues adjacent to tumours, 72 colorectal adenomas, and 716 normal colon tissues from healthy individuals) from 14 open access studies deposited in NCBI GEO and ArrayExpress. We calculated each sample's epigenetic age (EA) using eleven epigenetic clock models and derived the corresponding epigenetic age acceleration (EAA). For EA, we observed that most first- and second-generation epigenetic clocks reflect the chronological age in normal tissues adjacent to tumours and healthy individuals [e.g., Horvath (r = 0.77 and 0.79), Zhang elastic net (EN) (r = 0.70 and 0.73)] unlike the epigenetic mitotic clocks (EpiTOC, HypoClock, MiAge) (r < 0.3). For EAA, we used PhenoAge, Wu, and the above mitotic clocks and found them to have distinct distributions in different tissue types, particularly between normal colon tissues adjacent to tumours and cancerous tumours, as well as between normal colon tissues adjacent to tumours and normal colon tissue from healthy individuals. Finally, we harnessed these associations to develop a classifier using elastic net regression (with lasso and ridge regularisations) that predicts CRC diagnosis based on a patient's sex and EAAs calculated from histologically normal controls (i.e., normal colon tissues adjacent to tumours and normal colon tissue from healthy individuals). The classifier demonstrated good diagnostic potential with ROC-AUC = 0.886, which suggests that an EAA-based classifier trained on relevant data could become a tool to support diagnostic/prognostic decisions in CRC for clinical professionals. Our study also reemphasises the importance of open access clinical data for method development and training of young scientists. Obtaining the required approvals for controlled access data would not have been possible in the timeframe of this study.Copyright © 2023 Widayati, Schneider, Panteleeva, Chernysheva, Hrbkova, Beck, Voloshin and Chervova.