What it does
Developability Scout takes one or more binder sequences (VH/VL pair, scFv, VHH, or de novo scaffold) as FASTA or plain text and returns a 0 to 1 composite developability score plus a per-residue liability annotation. The composite rolls up five dimensions drawn from the published antibody developability literature and industry CMC practice: humanness, liability motifs, charge and pI, hydrophobic patches, and aggregation propensity.
Output includes per-residue heatmaps over the liability hotspots, CDR-level breakdowns, and side-by-side comparison across a full panel. CSV and PDF export for design reviews. A typical panel of 100 binders scores in seconds.
What it is not
Developability Scout is a sequence-level triage tool. It does not replace a full CMC developability assessment with HIC, SEC-MALS, PSR, and stability studies. It does catch the liabilities that are visible from sequence alone: deamidation hotspots (NG, NS, NT, NH), isomerisation sites (DG, DS, DH), unwanted N-linked glycosylation sequons (N-X-S/T) in variable regions, methionine oxidation sites, unpaired cysteines, extreme pI, and VH/VL charge asymmetry. Most of these should be detected and removed before a sequence ever reaches gene synthesis.
Methodology
Humanness is computed as germline similarity against IMGT references plus repertoire frequency against OAS. Liability motifs use position-aware regex against the CDR regions as defined by IMGT numbering. Charge uses theoretical pI and net charge at physiological pH with VH/VL asymmetry as a separate feature. Hydrophobic patches use Kyte-Doolittle sliding-window hydrophobicity weighted by CDR position. Aggregation propensity uses published per-residue propensity scales with aggregation-prone region (APR) detection for contiguous hydrophobic and beta-sheet-prone stretches. Weights are equal (20 percent each) by default and can be adjusted per campaign context.
Methodology is transparent: every score in the report links to the specific motif or residue stretch that drove it, not a black-box number.
When to use it
Between design and synthesis on a de novo campaign. The cheapest liability to remove is the one that never gets ordered. Run every RFdiffusion, BindCraft, or ProteinMPNN output through a developability pass before gene synthesis — a single glycosylation sequon introduced by ProteinMPNN sequence design is trivial to spot on paper and expensive to debug after expression.
Also useful for: triaging an in-house antibody panel before expression, checking naive yeast display hits before affinity maturation, and comparing VH/VL pairings on a single scorecard before freezing a clinical candidate.
Pricing
Free forever, no account required. Unlimited sequences, unlimited exports. Runs on CPU — no queue, no credits, no upsell.
From triage to developable binders
Developability Scout surfaces the liabilities. Removing them in-campaign is what the Binder Pilot and AI Binder Sprint programs do in our own pipeline: RFdiffusion, BindCraft, and Boltzgen outputs are pre-filtered with the same scoring logic, then validated on yeast display before lead nomination. If you want your binders scored and engineered, not just screened, the Binder Pilot is the entry point.