Caduceus: Bi-Directional Equivariant Long-Range DNA Sequence Modeling
This work addresses the problem of accurate long-range DNA sequence modeling for genomics researchers, offering a novel architecture that improves performance on tasks like variant effect prediction.
The authors tackled the challenge of modeling long-range interactions and reverse complementarity in genomic DNA sequences by proposing Caduceus, a family of bi-directional, equivariant long-range DNA language models, which outperformed previous models on downstream benchmarks, including exceeding the performance of 10x larger models on a variant effect prediction task.
Large-scale sequence modeling has sparked rapid advances that now extend into biology and genomics. However, modeling genomic sequences introduces challenges such as the need to model long-range token interactions, the effects of upstream and downstream regions of the genome, and the reverse complementarity (RC) of DNA. Here, we propose an architecture motivated by these challenges that builds off the long-range Mamba block, and extends it to a BiMamba component that supports bi-directionality, and to a MambaDNA block that additionally supports RC equivariance. We use MambaDNA as the basis of Caduceus, the first family of RC equivariant bi-directional long-range DNA language models, and we introduce pre-training and fine-tuning strategies that yield Caduceus DNA foundation models. Caduceus outperforms previous long-range models on downstream benchmarks; on a challenging long-range variant effect prediction task, Caduceus exceeds the performance of 10x larger models that do not leverage bi-directionality or equivariance.