Rethinking Genomic Modeling Through Optical Character Recognition

2026-02-02

AI Summary

This work addresses the limitations of traditional genomic modeling, which treats DNA as a one-dimensional sequence and struggles to efficiently compress long-range context while often wasting computational resources in low-information regions. For the first time, the authors introduce an optical character recognition (OCR)-inspired paradigm into genomics, proposing a visual DNA encoder coupled with a document decoder architecture. By rendering DNA sequences into structured visual layouts, the model generates compact, reconstructable visual tokens that enable prompt-driven tasks such as reading, region localization, subsequence retrieval, and masked completion. Evaluated on sequences up to 450 kilobases, the approach achieves significantly higher fidelity than baseline models—despite using only 256,000 trainable parameters and nearly 20× fewer effective tokens—outperforming models with 985× more parameters while maintaining low computational overhead.

Technology Category

• Computer Vision: Diffusion Models for Vision
• Machine Learning: Deep Generative Models & Autoencoders
• Search and Optimization: Evolutionary Computation

Application Category

• Economics, Online Markets and Human Computation: Economic ramifications for generative AI infrastructure and applications
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

Abstract

Recent genomic foundation models largely adopt large language model architectures that treat DNA as a one-dimensional token sequence. However, exhaustive sequential reading is structurally misaligned with sparse and discontinuous genomic semantics, leading to wasted computation on low-information background and preventing understanding-driven compression for long contexts. Here, we present OpticalDNA, a vision-based framework that reframes genomic modeling as Optical Character Recognition (OCR)-style document understanding. OpticalDNA renders DNA into structured visual layouts and trains an OCR-capable vision--language model with a \emph{visual DNA encoder} and a \emph{document decoder}, where the encoder produces compact, reconstructible visual tokens for high-fidelity compression. Building on this representation, OpticalDNA defines prompt-conditioned objectives over core genomic primitives-reading, region grounding, subsequence retrieval, and masked span completion-thereby learning layout-aware DNA representations that retain fine-grained genomic information under a reduced effective token budget. Across diverse genomic benchmarks, OpticalDNA consistently outperforms recent baselines; on sequences up to 450k bases, it achieves the best overall performance with nearly $20\times$ fewer effective tokens, and surpasses models with up to $985\times$ more activated parameters while tuning only 256k \emph{trainable} parameters.

Problem

• genomic modeling
• sequence compression
• sparse semantics
• long-context understanding
• computational efficiency

Innovation

• OpticalDNA
• vision-based genomic modeling
• OCR-style document understanding
• visual DNA encoder
• high-fidelity compression