Ranomics
ColabFold AlphaFold2 predicted protein structure colored by per-residue pLDDT confidence score
Structure prediction

ColabFold fast protein structure prediction

The Ranomics ColabFold tool runs a fast single-sequence fold: AlphaFold2 weights with no MSA and no templates, built for high-throughput triage of designed binders and orphan sequences, where an MSA adds little. For MSA-based folding of natural proteins, use the AlphaFold2 tool.

Monomer or multimer, single-sequence speed, seconds per model. Per-residue pLDDT and PAE on every output.

Hosted at tools.ranomics.com. Funded from your wallet, with $5 free to start.

How it works

From sequence to folded model in one pass

01

Submit sequence

Paste a FASTA sequence or upload a multi-sequence file. Monomers run as-is; multimers use a colon to separate chains.

02

Single-sequence fold

The hosted tool folds from the query sequence alone, no MSA search. That is the speed win for designed and orphan sequences, which carry no meaningful evolutionary alignment anyway.

03

AF2 inference

Run the AlphaFold2 weights on the single sequence with configurable recycle iterations. No MSA and no templates by default.

04

Folded model

Download a ranked PDB ensemble with per-residue pLDDT, PAE matrices, and predicted TM-score for multimer interfaces.

Methodology

What ColabFold changes inside the AF2 pipeline

The hosted ColabFold tool keeps the AlphaFold2 weights unchanged and skips the MSA search entirely, folding from the single query sequence. The throughput gain is the point: seconds per model for triaging large designed sets. When an MSA matters, the AlphaFold2 tool runs ColabFold with a full MMseqs2 search.

No MSA

Single-sequence folding

The hosted tool runs colabfold_batch in single-sequence mode, no homology search. Fastest path to a fold when the sequence is designed or orphan and an MSA would add noise, not signal.

Monomer or multimer

Complex prediction

Fold a monomer, or separate chains with a colon to predict a complex and get ipTM at the interface, all in single-sequence mode.

pLDDT

Per-residue confidence

pLDDT 0-100 per residue tells you which regions of the model to trust. PAE matrix tells you which inter-residue distances are confidently resolved.

AF2 weights

Same AlphaFold2 network

Identical AlphaFold2 weights as the full pipeline. Only the MSA is dropped, so a good single-sequence fold is a genuine AF2 prediction, just faster and with less context.

3 default

Recycle iterations

Each recycle feeds the previous prediction back as input. Default 3, configurable up to 12, with diminishing returns past 6 for most targets.

Beyond a single fold

Multimer, model choice, and batch throughput

Multimer mode

Predicts protein complexes using AF2-multimer weights or the residue-index-jump trick on the monomer model. Returns a predicted interface TM-score and pAE matrix so you can quantify confidence in the inter-chain contacts, not just the chain folds.

When to pick which fold

ColabFold for throughput on natural sequences with MSAs. Full-MSA AF2 for the final, highest-accuracy fold on a single chosen design. ESMFold for the fastest single-sequence pass on designed binders, orphan sequences, or anything where an MSA would not help.

Batch throughput

ColabFold is the throughput tool of choice for triaging large design pools. Fold an entire RFdiffusion or BindCraft output set in a single sitting, then promote the highest-confidence handful to full-MSA AF2 or wet-lab validation.

When to use ColabFold

When speed beats the last decimal of accuracy

Full-MSA AlphaFold2 is the gold standard for a single, definitive structure. But most real campaigns generate hundreds or thousands of candidate sequences, and you cannot afford to fold them all with the slow path. ColabFold is the throughput tool that bridges the gap between sequence generation and final-model validation.

Use ColabFold to rank a design pool, screen mutants, or survey a target family. Promote the top handful to full-MSA AF2 or experimental validation.

Triaging a pool of designed binder sequences before committing to wet-lab validation

Folding all variants in a deep mutational scanning library to flag misfolded designs

Surveying a target protein family by folding all orthologs in a single pass

Running a design-validation pass on RFdiffusion, BindCraft, or BoltzGen outputs

Building a structural prior for downstream docking, MD, or interface analysis

Quick sanity-check folds where 1-2 angstrom accuracy is enough for the decision

Fold your first sequence today

Create a free account on tools.ranomics.com and run ColabFold on a designed sequence. No shared queues, no install.

Method: Mirdita et al., ColabFold: making protein folding accessible to all. Nat Methods 19, 679-682 (2022). See also all Ranomics technology.