BioPhi quickstart
Call biophi_humanness for an OASis score and class, or biophi_humanize and biophi_graft to return a humanized VH/VL with the mutations made.
You have an antibody VH/VL pair and you want to know how human it is — and, if it isn't human enough, a humanized version that keeps the CDRs. BioPhi does both: paste the sequences, pick a mode, submit, and read the OASis score or the humanized sequence off the result card.
Score or humanize
Provide VH (and VL)
Paste the heavy-chain variable-region sequence, and the light chain too if you have it. Or name a known therapeutic and let the platform resolve the VH/VL from Thera-SAbDab.
Pick which tool to call
Call biophi_humanness to only score, biophi_humanize to rewrite the sequence with the Sapiens model, or biophi_graft to humanize by CDR grafting — the mode follows from which one you call. For humanization, set a target humanness; biophi_humanize also lets you toggle whether known binding residues stay fixed.
Submit and wait
BioPhi runs as an asynchronous CPU job on Modal. Submit it, and a result card appears on the canvas when the job finishes.
Read the result
In scoring mode the card shows the OASis humanness score, percentile, and human/humanized/chimeric/murine class, broken down per CDR and framework. In humanization mode it also shows the humanized VH/VL and the mutations made.
In chat
Describe the antibody and what you want done — the model routes to BioPhi and chains numbering, repertoire mining, or stability around it.
What to do next
- Confirm manufacturability — run antibody developability on the humanized candidate, since humanization can shift charge and aggregation.
- Number the humanized sequence — run ANARCI to re-annotate CDRs.
- Cross-check with a second metric — score the same sequence with Humatch.
Read next
Was this page helpful?
BioPhi outputs and interpretation
Humanness scores, classification, humanized sequences, and mutations from BioPhi.
How BioPhi results compare to the bench
What an OASis score and a Sapiens-humanized sequence have been shown to agree with experimentally, and why humanness does not predict immunogenicity.