How to Use BindFilter

BindFilter is a binder evaluation workflow for screening many candidate proteins, antibodies, nanobodies, or peptides against a target. Rather than generating binders itself, it co-folds each target-binder pair and aggregates orthogonal evidence spanning interface quality, predicted affinity, structural energetics, solubility, stability, aggregation risk, toxicity, and optional mutation-scanning readouts.

On Neurosnap, researchers provide Target Sequences together with a library of Input Sequences, choose a folding backend, and enable only the analysis modules relevant to the campaign. That makes BindFilter useful after de novo design, affinity maturation, or sequence-library generation when the main question is which candidates deserve experimental follow-up.

BindFilter is best understood as a Neurosnap orchestration layer around complementary structure- and sequence-based models. Each candidate binder is folded against the target with either a Chai or Boltz backend, after which the pipeline can apply methods such as Prodigy for affinity estimation, EvoEF2 for energetic scoring, NetSolP for solubility, TemStaPro for thermostability, Aggrescan3D for aggregation propensity, ToxinPred for peptide toxicity, and optional mutation-scanning models. The goal is not a single score, but a panel of partially independent signals that helps separate promising binders from candidates that only look good under one metric.

On Neurosnap, Fold Backend, MSA Mode, recycling, and diffusion settings control the structure-generation step, while the tool-selection toggles determine how broad the downstream screen will be. Results are organized for candidate comparison: interface-focused metrics help judge whether a binder engages the intended target, and developability or energetics summaries highlight liabilities that could block expression, formulation, or optimization.

• Study-fit inputs: BindFilter starts from target sequences and a binder library, which matches the stage where many designed or selected candidates need to be compared against the same target.
• Orthogonal evidence: The workflow combines co-folding, affinity proxies, energetics, solubility, stability, aggregation, toxicity, and optional mutation scanning so candidates are not ranked by a single fragile metric.
• Configurable screening: Researchers can choose the folding backend, adjust sampling behavior, and switch individual analysis modules on or off to match the campaign and binder modality.
• Research workflow fit: Running the folding and downstream filters in one place reduces the manual work of reconciling separate tools before selecting candidates for experiments.