DNALONGBENCH: a benchmark suite for long-range DNA prediction tasks.

Modeling long-range DNA dependencies is crucial for understanding genome structure and function across diverse biological contexts. However, effectively capturing these dependencies, which may span millions of base pairs in tasks such as three-dimensional (3D) chromatin folding prediction, remains a major challenge. A comprehensive benchmark suite for evaluating tasks that rely on long-range dependencies is notably absent. To address this gap, we introduce DNALONGBENCH, a benchmark dataset covering five key genomics tasks with long-range dependencies up to 1 million base pairs: enhancer-target gene interaction, expression quantitative trait loci, 3D genome organization, regulatory sequence activity, and transcription initiation signals. We assess DNALONGBENCH using five methods: a task-specific expert model, a convolutional neural network (CNN)-based model, and three fine-tuned DNA foundation models - HyenaDNA, Caduceus-Ph, and Caduceus-PS. We envision DNALONGBENCH as a standardized resource to enable comprehensive comparisons and rigorous evaluations of emerging DNA sequence-based deep learning models that account for long-range dependencies.