Towards A Generative Protein Evolution Machine with DPLM-Evo

Proteins are shaped by gradual evolution under biophysical and functional constraints. Protein language models learn rich evolutionary constraints from large-scale sequences, and discrete diffusion-based protein language models (e.g., DPLMs) are promising for both understanding and generation. However, existing DPLMs typically rely on masked diffusion that contradicts a simple biological intuition: proteins evolve through accumulated edits, not by emerging from masks. Consequently, these frameworks lack explicit pretraining objectives for substitution and insertion/deletion (indel) operations, limiting both optimization-style post-editing and flexible guided generation. To address these limitations, we present DPLM-Evo, an evolutionary discrete diffusion framework that explicitly predicts substitution, insertion, and deletion operations during denoising. DPLM-Evo decouples an upsampled-length latent alignment space from the variable-length observed sequence space, which makes indel-aware generation tractable. To better align substitutions with real evolution, we further introduce a contextualized evolutionary noising kernel that produces biologically informed, context-dependent mutation patterns. Across tasks, DPLM-Evo improves sequence understanding and achieves state-of-the-art mutation effect prediction performance on ProteinGym in the single-sequence setting. It also enables variable-length simulated evolution, and post-editing/optimization of existing proteins via explicit edit trajectories.

Proteins are the tiny machines that carry out many essential tasks in living cells. Designing useful proteins, such as better enzymes or therapeutic molecules, often requires changing an existing protein while keeping its shape and function intact. In nature, proteins evolve through small edits to their amino acid sequences: some amino acids are replaced, some are inserted, and some are deleted. This paper introduces an AI system that learns to make these kinds of evolutionary edits. Instead of only filling in missing parts of a fixed-length protein sequence, the system can change both the content and the length of a protein, making it closer to how real evolution works. It learns from large collections of natural protein sequences and uses this knowledge to judge mutations, generate new protein variants, and improve existing proteins. Our results show that this approach can better predict the effects of protein mutations and can create diverse protein sequences that are still likely to fold into valid structures. This could help researchers explore protein design more efficiently.