Development and Validation of an Automated DNA-Encoded Library Screening Data Analysis Platform: PB-DEL Autoscreening Analysis (PB-DELASA)

Tools
available for analyzing next-generation sequencing (NGS)
data produced from DNA-encoded library (DEL) screening campaigns are
often constrained to customized methods developed internally by individual
institutes, which usually generate data specifically focusing on protein–ligand
interactions and based on distinguished criteria of compound recommendation.
Existing approaches do not consider sequencing depth, sequencing error,
and quality control when identifying candidate compounds. The analysis
processes and criteria of compound recommendation for off-DNA synthesis
and confirmation are highly time-consuming and subjective, significantly
hindering the application of DEL screening in novel drug discovery.
Here, to address these challenges, we developed an integral, accurate,
and automated analysis workflow containing the tractability of the
building blocks and DNA tags in split-and-pool cycles, 2D and 3D plots,
and an enriched compound list, which was constructed based on computational
analysis, artificial intelligence, and the experiential knowledge
of medicinal chemists. This automated and standardized workflow was
further validated through a showcase screening campaign on a novel
antitumor target of CDK9. Novel hit compounds with high potency and
selectivity were identified efficiently with minimal synthesis effort.
The source code is available at https://github.com/kelly1210/PB-DELASA.

当前用于分析由DNA编码文库（DEL）筛选实验产生的下一代测序（NGS）数据的分析工具，往往局限于各研究机构自主开发的定制化方法；这类方法通常仅聚焦于蛋白质-配体相互作用的数据产出，且基于独特的化合物推荐标准。现有方法在筛选候选化合物时，并未纳入测序深度、测序错误及质量控制等关键考量因素。针对脱离DNA的化合物合成与验证环节的化合物推荐分析流程及标准，不仅耗时极长且主观性较强，这极大阻碍了DEL筛选在创新药物研发中的应用。为此，本研究针对上述痛点开发了一套完整、精准且自动化的分析流程，该流程可处理拆分-汇集循环中的结构单元与DNA标签数据，支持生成二维、三维可视化图表，并输出富集化合物列表；本流程基于计算分析、人工智能及药物化学家的经验知识构建而成。这套自动化标准化流程通过针对新型抗肿瘤靶点CDK9的示范性筛选实验得到了验证。研究人员仅需投入极少的合成工作量，即可高效筛选得到具有高活性与选择性的新型命中化合物。本研究的源代码可于https://github.com/kelly1210/PB-DELASA获取。