The central dogma of molecular biology outlines the flow of genetic information from DNA to RNA to proteins. With a limited vocabulary of four nucleotides (A, C, G, and T), DNA encodes an extensive instruction set, including the chromosomal positions where RNAs begin to be transcribed and the magnitudes of their expression. This process, known as transcription initiation (1), begins at transcription start sites (TSSs) and depends on RNA polymerase II recruitment to promoter sequences (2). However, the sequences and rules that govern transcription initiation remain elusive. On page 405 of this issue, Dudnyk et al. (3) use an explainable deep-learning model to find a small set of DNA sequence motifs that predict the position and activity of most TSSs in the human genome. The findings define a set of rules that govern transcription initiation and highlight the potential of using deep-learning approaches to understand how information is genetically encoded.

中文翻译：

---

DNA如何编码转录起始点

分子生物学的中心法则概述了遗传信息从 DNA 到 RNA 再到蛋白质的流动。 DNA 的词汇量有限，只有四种核苷酸（A、C、G 和 T），可编码广泛的指令集，包括 RNA 开始转录的染色体位置及其表达量。这一过程称为转录起始 ( 1 )，从转录起始位点 (TSS) 开始，并依赖于 RNA 聚合酶 II 募集至启动子序列 ( 2 )。然而，控制转录起始的序列和规则仍然难以捉摸。在本期第 405 页，Dudnyk等人。 ( 3 ) 使用可解释的深度学习模型来寻找一小组 DNA 序列基序，这些基序可以预测大多数 TSS 在人类基因组中的位置和活性。研究结果定义了一套控制转录起始的规则，并强调了使用深度学习方法来理解信息如何进行基因编码的潜力。