Evolve antibody sequences or score paired-antibody variants with a neural continuous-time Markov chain. Learn more
CoSiNE webserver overview
The CoSiNE webserver scores paired-antibody deep mutational scanning (DMS) variants with a neural continuous-time Markov chain. The available variant effect prediction (VEP) workflow returns CoSiNE log-likelihoods, perplexities, correlations with submitted experimental fitness, and optional Thrifty-corrected selection scores.
The interface also includes CoSiNE's unguided and predictor-guided antibody evolution controls. Evolution jobs are currently unavailable because the source project has not published the required RefineGNN binding-predictor checkpoint. See How to use CoSiNE online for a complete VEP workflow, result interpretation, validation, and the current evolution boundary.
Pricing
A CoSiNE submission currently has a fixed quote of 100 credits. The exact quote is shown before submission. Input size, scoring axes, branch lengths, and other settings do not currently change this quote.
Inputs
Variant effect prediction
| Input | Format | Required | Requirements |
|---|---|---|---|
| Paired-antibody DMS table | CSV, up to 10 MiB | Yes | One table with heavy-chain amino-acid sequences, light-chain amino-acid sequences, and numeric experimental fitness. All heavy sequences must have one consistent length, and all light sequences must have one consistent length. |
| Wild-type heavy-chain coding sequence | FASTA or plain DNA, up to 1 MiB | Conditional | Required when Thrifty correction is enabled for heavy or combined scoring. Must contain only A, C, G, and T and translate exactly to the resolved heavy-chain consensus. |
| Wild-type light-chain coding sequence | FASTA or plain DNA, up to 1 MiB | Conditional | Required when Thrifty correction is enabled for light or combined scoring. Must contain only A, C, G, and T and translate exactly to the resolved light-chain consensus. |
The default DMS column names are heavy, light, and fitness. Map different column names in DMS column mapping. CoSiNE derives the wild-type heavy and light consensus site by site unless you provide explicit consensus sequences.
Antibody evolution
| Input | Format | Required | Requirements |
|---|---|---|---|
| Starting antibody sequence | FASTA or plain protein, up to 1 MiB | No | One heavy-chain variable-domain sequence using the 20 canonical amino acids. When omitted, CoSiNE selects a source-provided binder using Built-in binder index. |
Antibody evolution is not currently executable because its required source checkpoint is unpublished.
Settings
Task
| Parameter | Options | Default | Description |
|---|---|---|---|
| Task | Antibody evolution, Variant effect prediction | Antibody evolution | Selects the input form, settings, and outputs. Choose Variant effect prediction for the currently runnable workflow. |
Variant effect settings
| Parameter | Type | Default | Description |
|---|---|---|---|
| CoSiNE branch lengths | Comma-separated non-negative numbers | 0.2 | Evaluates every requested CTMC branch length and adds ll_<t> and ppl_<t> columns. |
| Scoring axes | One or more of heavy, light, combined | heavy, light | Scores variants with mutations on the selected chain or on the paired heavy-light sequence. |
| Use paired heavy-light context | Boolean | true | For heavy or light, supplies both wild-type chains as context while changing the selected chain. combined always uses both chains. |
| Apply Thrifty SHM correction | Boolean | true | Subtracts a neutral somatic-hypermutation log-likelihood from the CoSiNE log-likelihood. Requires coding DNA for every selected chain. |
| Thrifty branch lengths | Comma-separated non-negative numbers | 0.2 | Evaluates every requested neutral SHM branch length and adds corrected_ll_<t>_<bl> columns. Visible when correction is enabled. |
| Correlation method | Spearman, Pearson | Spearman | Correlates each score column with the submitted experimental fitness values. |
| Log-transform experimental fitness | Boolean | false | Applies the natural logarithm to fitness before correlation. Use only when every analyzed fitness value is positive and the assay scale justifies the transform. |
DMS column mapping
| Parameter | Default | Description |
|---|---|---|
| Heavy-chain column | heavy | CSV column containing aligned-length heavy-chain amino-acid sequences. |
| Light-chain column | light | CSV column containing aligned-length light-chain amino-acid sequences. |
| Experimental fitness column | fitness | CSV column containing numeric assay values. |
| Heavy mutation column | Empty | Optional precomputed heavy-chain mutation labels, such as A23V or A23V,T45G. When empty, labels are derived from the sequences. |
| Light mutation column | Empty | Optional precomputed light-chain mutation labels. When empty, labels are derived from the sequences. |
| Heavy-chain consensus | Empty | Optional explicit wild-type heavy-chain amino-acid sequence. When empty, CoSiNE resolves a site-by-site consensus from the table. |
| Light-chain consensus | Empty | Optional explicit wild-type light-chain amino-acid sequence. When empty, CoSiNE resolves a site-by-site consensus from the table. |
Antibody evolution settings
These settings reproduce the source evolution interface, but evolution jobs remain unavailable until the required binding-predictor checkpoint can be verified.
| Parameter | Type | Default | Description |
|---|---|---|---|
| Sampling method | Unguided Gillespie, Guided Gillespie (TAG) | Unguided Gillespie | Uses unconditional sampling or predictor-guided sampling. |
| Binding predictor | SARS-CoV-2 Beta RBD, SARS-CoV-1 RBD | SARS-CoV-2 Beta RBD | Chooses the source predictor used for guided sampling and native fitness columns. |
| Built-in binder index | Integer, minimum 0 | 0 | Zero-indexed binder selected from the chosen predictor dataset when no starting sequence is supplied. |
| Branch length | Number, minimum 0 | 2 | Evolutionary time calibrated to the expected number of mutations per site. |
| Number of trajectories | Integer, minimum 1 | 100 | Number of sampled evolutionary trajectories. |
| Use discrete mutation steps | Boolean | false | Uses Maximum mutations as the discrete Gillespie step count instead of continuous branch time. |
| Maximum mutations | Integer, minimum 1 | Empty | Optional mutation ceiling in continuous mode and required step count in discrete mode. |
| Mutable region | All positions, CDR1, CDR2, CDR3, All CDRs, Framework 1, Framework 2, Framework 3, Framework 4, All framework regions | All positions | Restricts changes to the selected IMGT-numbered region. |
| Guidance strength | Number | 2 | Predictor-guidance strength. Visible only for guided sampling. |
| Random seed | Integer from 0 to 2147483647 | Empty | Reproduces sampled trajectories when set. An empty value uses the source timestamp seed. |
Outputs
Variant effect prediction
| Artifact | Format | Description |
|---|---|---|
| Axis result tables | CSV | cosine_vep_heavy.csv, cosine_vep_light.csv, and/or cosine_vep_combined.csv, one for each selected axis. They preserve source row order and include the submitted columns, mutation labels, raw scores, and optional corrected scores. |
| Raw correlations | CSV | cosine_correlations.csv with chain, metric, correlation, p_value, and n_samples. |
| Selection-score correlations | CSV | cosine_selection_correlations.csv, returned when SHM correction is enabled, with one row per chain and correction branch length and one column per CoSiNE branch length. |
| Submitted DMS table | CSV | submitted_dms.csv, preserved unchanged. |
| Source log | TXT | cosine_source.log with captured source diagnostics. |
| Provenance | JSON | cosine_provenance.json with pinned source, model, compatibility, and resolved-setting details. |
Antibody evolution
When evolution becomes available, its interface returns the native evolution table as cosine_evolution.source.csv, a derived cosine_evolved_sequences.fasta, and cosine_source.log. Planned table fields include the seed and sampled sequences, branch length, guidance type, mutation count, trajectory index, timing, predictor fitness, humanness metrics, and edit distance from the seed. No evolution result is produced while the required source checkpoint is unavailable.
Understanding results
mutationidentifies the amino-acid substitutions scored on that axis. Combined labels join heavy and light mutations with|.ll_<t>is the summed CoSiNE log-likelihood for the wild-type-to-variant transition at branch lengtht. Within the same axis, context, and branch length, a higher value means the model assigns the transition greater likelihood.ppl_<t>isexp(-ll / sequence length). Lower values indicate lower per-residue surprise. Because sequence length and context affect the calculation, compare perplexity within a consistent scoring setup.corrected_ll_<t>_<bl>is the CoSiNE log-likelihood minus the Thrifty neutral SHM log-likelihood. It is a model-derived selection score, not a measured affinity or binding constant.correlationis the selected Spearman or Pearson association between one score column and the submitted fitness values. Its sign depends on the metric: likelihood-based scores and perplexity point in opposite numerical directions.p_valueandn_samplesdescribe that correlation calculation. They do not provide a universal model confidence score.
Use CoSiNE scores to prioritize and compare variants within a controlled assay and scoring protocol. Confirm candidates with binding, expression, developability, and other relevant experiments.
