Innate immune receptor screening, without the wait.
Predict binding strength for any of 14 innate immune receptors in under 2 seconds. Stop running AutoDock Vina manually - filter your entire candidate library before you synthesize a single compound.
R²=0.71
ML model accuracy
14
Receptors covered
<2s
Per analysis
$0
To start
Comparison
What ImmunoBind replaces
Traditional workflow
AutoDock Vina: 2-4 hours of setup per compound
Synthesis required before you know if a molecule binds
ELISA / NF-κB reporter assays: 3-5 days + ~$200 per compound
Manual pharmacophore analysis against crystal structures
Separate expert and toolchain needed for each receptor family
No consistent cross-receptor scoring or comparison
vs
With ImmunoBind
Binding estimate in under 2 seconds for any compound
Screen entire libraries before committing to synthesis
20 free queries per month - no credit card required
Automated pharmacophore report with PDB key contact residues
14 receptors in a single interface - TLR, STING, RIG-I, MDA5
Consistent pIC50 and IC50 output across all targets
Workflow
Three steps to a binding report
Step 01
Draw or paste your compound
Enter a SMILES string directly, use the built-in JSME structure editor, or search by compound name - PubChem lookup is built in and resolves common drug names and research compounds instantly.
→
Step 02
Select your target receptor
Choose from 14 innate immune sensors grouped by receptor family. TLR1/2, TLR2/6, and TLR4 use a trained XGBoost ML model (R²=0.71); all others use expert LLM analysis with receptor-specific pharmacophore context from published crystal structures.
→
Step 03
Get your binding report
Receive a binding score (0-100), estimated pIC50 and IC50 in nM, key pharmacophore contact flags from PDB crystal structures, drug-likeness analysis, and a full expert narrative. Export as JSON or CSV, or share via link.
Receptor coverage
Every major innate immune sensor
TLR1/2 · TLR2/6 · TLR4 - ML model
The three most drug-targeted TLRs. Binding predictions use a trained XGBoost model (R²=0.71) fit on experimental pIC50 data from published lipopeptide and LPS analogue studies. Outputs include estimated IC50 in nM, pharmacophore contact scores for key residues (Asn294, Phe349, Glu271), and lipid anchor detection. The model takes 2D molecular descriptors: MW, LogP, acyl chain count, and fingerprint-derived features.
TLR3 · TLR7 · TLR8 · TLR9 - nucleic acid sensors
Endosomal TLRs that detect nucleic acid patterns from pathogens. TLR3 recognises double-stranded RNA (poly I:C analogues; PDB 3CIY); TLR7 and TLR8 sense single-stranded RNA and small-molecule agonists such as imiquimod and resiquimod (PDB 5GMH, 3W3J); TLR9 detects unmethylated CpG ODNs from bacteria and viruses (PDB 3WPB). Each analysis is grounded in the corresponding crystal structure.
STING · cGAS - cytosolic DNA sensing
STING (Stimulator of Interferon Genes) and its upstream sensor cGAS detect foreign or misplaced cytosolic DNA - a hallmark of viral infection, cancer mutation load, or cellular stress. STING agonists (cyclic dinucleotides, DMXAA analogues; PDB 4LOH) are among the most active areas in cancer immunotherapy, with multiple compounds in Phase II/III trials. cGAS inhibitors are being explored in autoimmune disease.
RIG-I · MDA5 - cytosolic RNA sensing
DExD/H-box RNA helicases that sense double-stranded RNA generated during viral replication. RIG-I preferentially recognises short 5′-triphosphate dsRNA (PDB 4A36); MDA5 forms cooperative filaments on long dsRNA (PDB 4GL2). Both activate the MAVS signalling axis, triggering type-I interferon production. Active drug programmes target influenza, SARS-CoV-2, and RSV using synthetic RIG-I / MDA5 agonists.
Under the hood
How the predictions are made
ML model - TLR1/2, TLR2/6, TLR4
XGBoost · R²=0.71
Trained on experimental pIC50 measurements from published lipopeptide binding studies. Features include molecular weight, cLogP, rotatable bond count, acyl chain count, hydrogen bond donors and acceptors, and 2048-bit Morgan fingerprint-derived descriptors. The model outputs a pIC50 value directly; IC50 is back-calculated as 10^(9−pIC50) nM. Confidence is highest for lipopeptide scaffolds structurally similar to training data (Pam3CSK4, FSL-1, lipid A analogues).
LLM analysis - all other receptors
Claude Sonnet · Expert context
Each of the 14 receptors has a hand-curated pharmacophore context: key binding residues from X-ray crystallography (PDB), known agonist and antagonist scaffolds with their potency ranges, and selectivity-determining structural features. The LLM evaluates the input SMILES against this expert context and returns a structured binding score (0-100), pharmacophore match score, key contact flags, drug-likeness assessment, and a free-text expert narrative explaining the rationale.
Use cases
Built for researchers who move fast
💉
Vaccine adjuvant design
Screen synthetic lipopeptide and TLR agonist libraries before committing to SPPS or total synthesis. Rapidly distinguish TLR1/2 vs TLR2/6 selectivity - the two heterodimers have distinct binding pocket geometries and acyl chain preferences that ImmunoBind resolves in a single screen.
🧬
Cancer immunotherapy
Evaluate STING agonists, TLR7/8 agonists, and RIG-I activators for tumour microenvironment reprogramming. Screen cyclic dinucleotide analogues, DMXAA derivatives, and imidazoquinoline scaffolds against the appropriate receptor before moving into cell-based NF-κB or IFN-β reporter assays.
🦖
Antiviral drug discovery
Identify compounds that activate innate antiviral responses via RIG-I, MDA5, TLR3, TLR7, or TLR9. Useful for screening candidates that trigger interferon responses as part of a combination antiviral strategy, or for identifying off-target TLR activation in drug safety profiling.
Enter a SMILES string above and select a receptor to analyze binding potential
Batch screening
Upload a CSV with a smiles column - up to 20 compounds per run.
↑
Drop CSV here or click to browse
Max 20 compounds · Required: smiles · Optional: name
Or try demo:
Analyzing...0 / 0
Total
0
High
0
Medium
0
Low
0
Results
Name
SMILES
pIC50
IC50 (nM)
MW
LogP
Verdict
Upload a CSV or load the demo to screen a compound library
Analysis history
Saved locally in your browser - never leaves your device
No analyses saved yet. Run a screener analysis and results will appear here automatically.
Benchmark validation
Five compounds with known experimental activity, run through ImmunoBind. Green = model agrees with literature; amber = partial match; red = disagreement. Click any row to run a live analysis.
What this tests
A screening tool is only as trustworthy as its ability to correctly rank compounds we already know the answer to. These five compounds span the receptor families covered by ImmunoBind, with published IC50 or EC50 values from peer-reviewed literature. The tool should score agonists high and negative controls low — if it doesn't, that's a signal to distrust a result.
Live result
Literature sources
Pam3CSK4 TLR1/2 EC50 ~0.1–1 ng/mL — Aliprantis et al. Science 1999; Jin et al. Cell 2007 (PDB 2Z7X).