Hold out the genome: a roadmap to solving the cis-regulatory code

Abstract

Gene expression is regulated by transcription factors (TFs) that work together to read cis-regulatory DNA sequences. The "cis-regulatory code" -how cells interpret DNA sequences to determine when, where, and how much genes should be expressed -has proven to be exceedingly complex 1,2 . Recently, advances in the scale and resolution of functional genomics assays and Machine Learning (ML) have enabled significant progress towards deciphering this code 3-6 . However, the cis-regulatory code will likely never be solved if models are trained only on genomic sequences; regions of homology can easily lead to overestimation of predictive performance, and our genome is too short and has insufficient sequence diversity to learn all relevant parameters. Fortunately, randomly synthesized DNA sequences enable testing a far larger sequence space than exists in our genomes, and designed DNA sequences enable targeted queries to maximally improve the models. Since the same biochemical principles are used to interpret DNA regardless of its source, models trained on these synthetic data can predict genomic activity, often better than genome-trained models 7,8 . Here, we provide an outlook on the field, and propose a roadmap towards solving the cis-regulatory code by a combination of ML and massively parallel assays using synthetic DNA. ContributionsCGD and JT conceptualized the paper. CGD produced the first draft, analyzed the data, and made the figures with advice from JT. CGD and JT edited the manuscript.

Plant Synthetic Promoters

“…The further growth of libraries above 10 8 elements requires the application of synthetic DNA fragments not related to known DNA genome sequences [46]. However, the analysis of such large libraries usually employs ML or DL to identify relations between DNA sequence properties and promoter activity [46,50,[241][242][243]251,262,263]. To improve these ML and DL approaches, they are trained on synthetic, random DNA fragments to test a larger sequence space; models trained on such synthetic data can predict genomic activity better than those solely trained on genome DNA [46,251].…”

“…However, the analysis of such large libraries usually employs ML or DL to identify relations between DNA sequence properties and promoter activity [46,50,[241][242][243]251,262,263]. To improve these ML and DL approaches, they are trained on synthetic, random DNA fragments to test a larger sequence space; models trained on such synthetic data can predict genomic activity better than those solely trained on genome DNA [46,251]. More complex relationships between synthetic promoter structure, activity and DNA methylation or histone acetylation status are efficiently addressed by multilayered DL algorithms, while the ML models could not grasp these labyrinthine qualities precisely [242,[247][248][249][250].…”

“…The mutual limitation of ML and DL model development is that available ML and DL models are developed and trained on genomic DNA that is too short, evolution-constrained and has insufficient sequence diversity to learn all relevant parameters. However, synthetic, random DNA sequences can be used to test and train a larger sequence space as compared to the genome: ML models trained on these synthetic data can predict genomic activity better than genome-trained models [46,251]. Moreover, the DL algorithms usually neglect evolutionary processes within biological systems, often resulting in false positives and counterfeit interpretations.…”

Epigenomic insights into common human disease pathology

“…In terms of distal CREs or enhancers, the regulatory role of DNAm is proposed to be highly context dependent [ 99 ]. Deciphering this cis -regulatory code is now being aided by ML and large-scale functional assays [ 100 ].…”

Improving the performance of supervised deep learning for regulatory genomics using phylogenetic augmentation

“…A major challenge in the field is determining how to train more complex deep learning models for applications outside of the most data-rich systems. A proposed solution is to substantially increase data volume by performing assays on randomly generated synthetic sequences, and then evaluating models trained on these sequences using true genomic sequences ( De Boer et al 2020 , de Boer and Taipale 2024 ). The reasoning behind this approach is that the genome does not contain sufficient variation to learn all aspects of the cis-regulatory code.…”