LineageFlow: Flow Matching for High-Fidelity Family-Aware Protein Sequence Generation

Protein sequence generation for engineering requires samples that are biophysically plausible and, when targeting a family/domain, remain recognizable members while exploring within-family diversity. Current discrete generative models typically start from uniform or masked-token noise, which discards strong position-specific constraints induced by evolution and forces the model to reconstruct conserved residues from scratch, leading to weak family control and low foldability. We propose \emph{LineageFlow}, a simplex-valued flow-matching model that initializes generation from lineage priors derived from ancestral sequence reconstruction, turning generation into structured mutation from an evolved scaffold. Across diverse protein families, LineageFlow achieves family validity close to held-out natural sequences and improves predicted structural confidence over uniform-/mask-initialized baselines while maintaining substantial within-family novelty and diversity, even surpassing a large pretrained baseline trained on substantially more data. Finally, we introduce \emph{rerouting}, a single intermediate-time mutate--select--amplify intervention that enables objective-guided sampling without per-step predictor guidance and yields further gains in plausibility, including a zero-shot enzyme generation case study.

LineageFlow is a method for generating new protein sequences that belong to a desired protein family. Protein families contain related proteins that often share structure and function, so being able to generate family-consistent sequences can help researchers explore useful variants for protein engineering. Instead of starting generation from random noise, LineageFlow begins from an evolutionary estimate of what an ancestral sequence in the target family may have looked like. This gives the model a family-specific starting point and helps it preserve important conserved positions. The method then gradually modifies this starting sequence into new candidate proteins, with an additional selection step that can steer the generated sequences toward more plausible designs. In experiments across thousands of protein families, LineageFlow generated sequences that were usually recognized as belonging to their intended families, while also producing diverse and novel candidates. We also tested the method on enzyme families and found that it could generate enzyme-like sequences that preserve important motifs and show promising predicted properties.