Logo image
Back
Evaluating Genomic Language Models: Experimental Validation, Multi-Omics Integration, and Ethical-Legal Challenges
Book chapter

Evaluating Genomic Language Models: Experimental Validation, Multi-Omics Integration, and Ethical-Legal Challenges

Mohammad Saleh Refahi, Hyunwoo Yoo, Gavin Hearne, Chaz Allegra, James R. Brown, Bahrad A. Sokhansanj, Stephen Woloszynek, Keyush Ramjattun, Shanmukha Rao Allipilli, Robi Polikar, …
Signal Processing and Biomedical Engineering Research, pp 89-134
02 Jul 2026
DOI: https://doi.org/10.1007/978-3-032-18509-9_3

Abstract

Data privacy and governance Ethics and policy Genomic language models Host–microbe interactions Metagenomics and microbiome
The rapid expansion of genomic data, driven by advancements in sequencing and multi-omics technologies, calls for analytical approaches that go beyond traditional methods to extract deeper biological insights. Genomic Language Models (GLMs) have emerged as powerful tools capable of integrating diverse omics data—including genomic, transcriptomic, epigenomic, proteomic, and metabolomic information—providing a comprehensive view of biological systems. By leveraging high-dimensional data across multiple biological scales, these models not only process large-scale sequence data but also uncover meaningful biological patterns, functional relationships, and novel regulatory mechanisms. In this work, we present a comprehensive review of GLMs, examining their core methodologies, tokenization strategies, and computational frameworks. We also conduct an experimental comparison of multiple GLMs to evaluate their efficiency, scalability, and generalizability across diverse genomic datasets. As GLMs integrate with multi-omics data, the need for models that not only process vast datasets but also extract meaningful biological insights becomes increasingly critical. The growing scale and fidelity of multi-omics data demand approaches capable of capturing complex relationships across genes, proteins, metabolites, and disease mechanisms. Deep learning models, including large language models (LLMs) and GLMs, hold promise in bridging molecular data with clinical insights, unlocking new opportunities in personalized medicine and biomedical discovery. While GLMs offer transformative potential, their application raises significant ethical and legal concerns, including data privacy, informed consent, proprietary data access, and biases in training datasets that may impact fairness and representation. Addressing these challenges is crucial to ensure the responsible and equitable use of genomic AI technologies.

Metrics

1 Record Views

Details

Logo image