BioMate runs AlphaFold 2 and AlphaFold 3 on AWS GPU instances. Submit a UniProt ID, FASTA sequence, or gene name — BioMate returns the predicted structure with pLDDT confidence scores, PAE plots, and a ready-to-dock PDB file, all in a single chat message.
BioMate automatically selects the appropriate AlphaFold version based on your target type. You can override the selection at any time from the parameter panel.
| Capability | AlphaFold 2 | AlphaFold 3 |
|---|---|---|
| Single-chain protein | Best accuracy; recommended for most targets | Supported; similar accuracy to AF2 |
| Protein complex (multimer) | AlphaFold-Multimer; good for homodimers and heterodimers | Improved multimer accuracy; use for large complexes |
| Protein–DNA / Protein–RNA | Not natively supported | Supported; transcription factor–DNA, ribosome–RNA |
| Protein–small molecule | Not supported | Supported; predict binding pose alongside structure |
| Compute requirement | Single A100 GPU; ~20–60 min for typical protein | Higher memory; ~45–120 min for complexes |
| pLDDT confidence | Per-residue (0–100) | Per-residue + ipTM for interface quality |
Per-residue confidence score from 0–100. Above 90: very high confidence (well-structured region). 70–90: confident. 50–70: low confidence (may be disordered). Below 50: likely intrinsically disordered.
2D matrix showing the expected positional error between residue pairs. Low PAE values across a domain indicate a well-defined relative arrangement. High PAE between domains indicates flexible or uncertain interdomain geometry.
AlphaFold 3 metric specific to complexes. Measures confidence in the predicted interface geometry. ipTM above 0.8 indicates a reliable complex prediction; below 0.5 suggests the interaction may not be well-defined.
| File | Format | Description |
|---|---|---|
| Predicted structure | PDB / mmCIF | 3D coordinates with B-factor column containing pLDDT per residue |
| Confidence JSON | JSON | Per-residue pLDDT array and summary statistics |
| PAE plot | PNG | Predicted aligned error heatmap; identifies domain boundaries and flexible regions |
| Multiple sequence alignment | A3M | MSA used as input; available for review or re-use with custom MSA |
| Methods report | DOCX | AlphaFold version, MSA depth, template hits, and all parameters cited for publication |
BioMate connects structure prediction directly to molecular docking. When you ask BioMate to "dock compound X against target Y," it automatically runs AlphaFold if no experimental structure exists, then feeds the high-confidence regions of the predicted structure into the docking pipeline — all in a single workflow run.
Yes. BioMate runs both AlphaFold 2 (single-chain proteins, high accuracy for most targets) and AlphaFold 3 (protein–DNA, protein–RNA, and protein–small molecule complexes). The version is selected based on your target type.
A FASTA sequence, UniProt ID, or gene name. BioMate fetches the sequence automatically from UniProt if you provide an ID.
pLDDT (predicted local distance difference test) ranges from 0–100. Scores above 90 indicate very high confidence; 70–90 are confident; below 50 indicate disordered or low-confidence regions. BioMate color-codes the structure by pLDDT and flags low-confidence binding site residues.
Yes. BioMate feeds AlphaFold output directly into AutoDock Vina or GLIDE in a single integrated pipeline, filtering out low-pLDDT regions from the docking grid automatically.
Provide a UniProt ID or FASTA sequence. BioMate runs AlphaFold on GPU, returns the structure with confidence annotations, and feeds it directly to docking if you need it.
Try free →