Decoding transcriptional regulation via a human gene expression predictor

Transcription factors (TFs) regulate cellular activities by controlling gene expression, but a predictive model describing how TFs quantitatively modulate human transcriptomes is lacking. We construct a universal human gene expression predictor named EXPLICIT-Human and utilize it to decode transcriptional regulation. Using the expression of 1613 TFs, the predictor reconstitutes highly accurate transcriptomes for samples derived from a wide range of tissues and conditions. The broad applicability of the predictor indicates that it recapitulates the quantitative relationships between TFs and target genes ubiquitous across tissues. Significant interacting TF-target gene pairs are extracted from the predictor and enable downstream inference of TF regulators for diverse pathways involved in development, immunity, metabolism, and stress response. A detailed analysis of the hematopoiesis process reveals an atlas of key TFs regulating the development of different hematopoietic cell lineages, and a portion of these TFs are conserved between humans and mice. The results demonstrate that our method is capable of delineating the TFs responsible for fate determination. Compared to other existing tools, EXPLICIT-Human shows a better performance in recovering the correct TF regulators. Copyright & COPY; 2023, The Authors. Institute of Genetics and Developmental Biology, Chinese Academy of Sciences, and Genetics Society of China. Published by Elsevier Limited and Science Press. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).

Automated summary

Researchers developed a universal human gene expression predictor called EXPLICIT-Human and used it to decode transcriptional regulation. The predictor accurately reconstructed transcriptomes for a wide range of tissues and conditions using the expression of 1613 transcription factors. It extracted significant interacting TF-target gene pairs, enabling inference of TF regulators for various pathways. The method showed better performance in recovering the correct TF regulators compared to existing tools.