Gene Regulatory Network Inference in the Presence of Selection Bias and Latent Confounders

Gongxu Luo, Haoyue Dai, Boyang Sun, Loka Li, Biwei Huang, Petar Stojanov, Kun Zhang

Keywords: gene regulatory network inference, selection bias, latent confounders, causal discovery

Abstract: The study of gene regulatory network inference (GRNI), with a focus on uncovering causal relations among genes, holds significant potential to explain fundamental biological processes, such as how cellular identity is established or disrupted in disease. Unfortunately, current methods fail to adequately interpret the widespread phenomena of differential gene expression. The limitation can largely be attributed to the overlook of the selection process (e.g., survival bias), which is ubiquitous and fundamental in biology. Furthermore, recent studies have shown that gene expression is regulated by latent confounders (e.g., non-coding RNAs). Both of which can lead to spurious dependencies, thereby distorting GRNI results. To mitigate these challenges, we propose a novel algorithm, called Gene Regulatory Network Inference in the presence of Selection bias and Latent confounders (GISL). It is designed to uncover the causal structure by leveraging data across multiple distributions obtained via gene perturbation. Surprisingly, we find that the qualitative structure information, selection process, and latent confounders are partially identifiable without any parametric assumption under mild graphical conditions. Experimental results on both synthetic and real-world single-cell gene expression datasets demonstrate the superiority of GISL over existing strong baseline methods.

Primary Area: causal reasoning

Code Of Ethics: I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics.

Submission Guidelines: I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2025/AuthorGuide.

Anonymous Url: I certify that there is no URL (e.g., github page) that could be used to find authors’ identity.

No Acknowledgement Section: I certify that there is no acknowledgement section in this submission for double blind review.

Submission Number: 7613