Library triage · binding affinity prediction

Rank compound libraries against 473 targets before synthesis, docking, or wet-lab screening.

A computational chemistry workbench — batch scoring, ChEMBL analog checks, medicinal chemistry filters, 3D pocket viewer, and CSV export. Prioritize compounds before expensive downstream work.

Start free Pathfinder demo

VectaBind workbench with VEGFR2 batch results, 3D pocket viewer, and compound #6 selected at pKd 7.53

1,000+

Compounds per batch score in the workbench

473+

Pre-computed targets across 19 therapeutic areas

80%

ChEMBL similarity on demo EGFR hit #6

Verify analogs in public databases

Evidence

Validated against known EGFR chemistry

Real scored output from the workbench — not a headline MAE. GEN-12 shows predicted pKd 8.06, 80% ChEMBL similarity to a published analog, QED 0.94, and Lipinski compliance. You can verify every signal.

EGFR hit #6 Strong binder

GEN-12 · Gen-EGFR

Predicted pKd8.06

QED0.94

ChEMBL similarity80%

LipinskiCompliant

Drug-like · high confidence · CHEMBL551595 analog

Open EGFR demo Full validation page

How predictions are grounded

①

Model prediction

Stage 6 EGNN affinity head outputs predicted pKd and binding probability per target pocket.

②

Medicinal chemistry filters

QED, Lipinski, LogP, and MPO weighting from RDKit — the same filters med chem uses daily.

③

Known chemistry

ChEMBL Tanimoto similarity surfaces analogs you can look up and compare in public databases.

Signal provenance

Signal	Source
pKd	VectaBind model
QED · Lipinski · LogP	RDKit
Similarity	ChEMBL
Pocket · 3D view	PDB structure
CSV export	Workbench session

Hit Triage Workbench

Screen libraries up to 1,000 compounds

Import HTS decks, generated sets, or SAR series — score the full list with progress, sortable results, MPO filters, SAR scatter, and CSV export. Compound IDs and series tags carry through from your library.

CSV upload with id, name, series, and SMILES columns
Chunked batch scoring with live progress (API tier sets request batch size)
Full results table, expand view, selectivity heatmaps, and 3D pocket overlay
Generated hits bridge back from REINVENT4 with GEN-* IDs preserved

Open workbench Batch scoring guide

1,000Max compounds per app batch score

<1sAverage latency per compound scored

500+Libraries stored in IndexedDB automatically

Workflow

Upload → filter → score → generate

Real protocol stages, not a product tour. Each step maps to a tab in the Hit Triage Workbench.

step 1

Upload library

Import CSV with SMILES, compound IDs, and series tags — up to 1,000 structures per library.

step 2

Filter by indication

Select targets by disease area — oncology, neuro, cardio — not PDB codes alone.

step 3

Batch score

Rank full libraries by pKd, MPO, and confidence. SAR scatter, selectivity heatmap, CSV export.

step 4

Generate & re-score

REINVENT4 RL with affinity as reward. Push hits back into your library and re-score.

What it does

hit triage

Filter libraries before you commit to synthesis

Upload a compound set (10 or 1,000 SMILES), select a target, and get a ranked affinity table in minutes. IDs, series tags, and confidence tiers surface automatically so you know which hits to advance.

EGFR · 10 cmp · batch scored

8.18HIGH

8.16HIGH

8.06HIGH

#10

7.67MED

REINVENT4 · CDK2 · 100 structures

GEN-0041 · 8.1 GEN-0082 · 7.9 GEN-0054 · 7.4 GEN-0013 · 6.2 GEN-0097 · 5.8

molecule generation

Generate and score in one loop

REINVENT4 integration runs goal-directed generation using the affinity model as the scoring oracle. Molecules are generated and scored in the same pipeline — no manual transfer steps.

pocket briefing + AI

3D pocket context and plain-language SAR

Interactive pocket viewer with contact labels. Pocket Briefing shows target metadata and selected compound structure. Ask binding, SAR, and drug-likeness questions without writing queries.

Pocket Briefing · VEGFR2 · Cmpd #1

Which hits have selectivity against VEGFR2 vs VEGFR1?

Cmpd #1 and #2 show >1.2 pKd selectivity for VEGFR2. Pocket contacts include CYS1045 and VAL899.

Model

Rank within your library — not absolute K_d

VectaBind is optimized for relative ranking within a library on a fixed target, not for predicting experimental K_d to two decimal places. See Validation for reproducible EGFR examples and ChEMBL-grounded spot-checks.

Rank ordering

Spearman ρ within a target — do stronger binders score higher than weaker ones in your series?

Within 1 log unit

% of predictions within 1.0 pKd of experimental values — practical tolerance for triage.

Top-k enrichment

Of the compounds you would synthesize first, how many assay active?

Internal PDBBind 2020 validation MAE: 0.20 pKd (model-selection split). Literature reference models and split details in Methods · Validation page.

Fit

Who this is for

Honest positioning — VectaBind is a fast screening layer, not a replacement for every downstream assay or simulation workflow.

Good fit

HTS deck triage before synthesis spend
SAR series ranking on a fixed target
Generate → score loops with REINVENT4
Academic exploration and method comparison
Small biotech teams without ML infrastructure

Not yet

Sole basis for IND or clinical go/no-go decisions
Replacing FEP or structure-based docking for lead optimization
Covalent binders, PROTACs, or macrocycles outside training distribution
Substitute for experimental K_i/K_d confirmation

Coverage

70+ scoreable targets · 473 in Pro

Pre-computed pocket embeddings for clinically relevant kinases, GPCRs, and enzymes.

Oncology

EGFR, KRAS, CDK4/6, BRAF, HER2, VEGFR2, MET, ALK + 70 more

Neurodegeneration

BACE1, MAPT, SNCA, LRRK2, APP + 30 more

Cardiovascular

ACC2, PCSK9, Factor Xa, Thrombin, ACE2 + 40 more

Inflammation

JAK1/2, TNF-α, IL-6R, COX-2, PDE4, BTK + 45 more

Infectious disease

SARS-CoV-2 Mpro, Influenza NA, HIV-1 PR + 35 more

+ 14 more areas

Metabolic, rare disease, mental health, liver, lung, bone, skin, eye

Pricing

Start free, scale as you grow

No credit card required for the free tier.

Free

^$0

no credit card

✓500 scores / month

✓Libraries up to 1,000 cmp

✓Hit Triage Workbench

✓70+ targets · API batch 10

Get started

Pro

^$299

per month

✓50,000 scores / month

✓Full 473-target catalog

✓API batch 1,000 / request

✓AI assistant + pocket viewer

Get Pro

Enterprise

custom volume

✓Unlimited scoring

✓Custom target onboarding

✓SLA + priority support

Talk to us