GLM-Prior: a genomic language model for transferable sequence-derived priors in gene regulatory network inference

Gene regulatory network inference depends on high-quality prior knowledge, yet curated priors are often incomplete or unavailable across species and cell types. We present GLM-Prior, a genomic language model fine-tuned to predict transcription factor-target gene interactions from nucleotide sequence. We integrate GLM-Prior with PMF-GRN in a dual-stage pipeline that combines sequence-derived priors with single-cell expression data for prior-conditioned GRN inference. Across six cell-line contexts, GLM-Prior performance scales with positive label abundance and TF coverage, showing above-chance agreement with reference networks in well-annotated mammalian settings. Single-species, species-transfer, and multi-species training show that GLM-Prior can construct informative priors across related mammalian species. Compared with accessibility-based priors, GLM-Prior achieves the highest prior performance in four of five mammalian cell lines. These benchmarks show that prior quality largely constrains GRN inference performance, positioning GLM-Prior as a transferable workflow for sequence-derived prior construction when matched experimental assays are unavailable.