Licensing¶
The mhctools license¶
mhctools uses the Apache License 2.0. It permits use, modification, and redistribution, including commercial use. You can include mhctools in an application without publishing that application's source code.
When redistributing mhctools or a modified version:
- Include a copy of the license.
- Retain applicable copyright and attribution notices, including notices from a supplied NOTICE file.
- Mark files you changed.
The license includes a limited contributor patent grant. It does not grant trademark rights, and the software comes without warranties. The full terms and Apache licensing FAQ explain these conditions.
Predictor licenses¶
Each upstream program, model, or dataset has its own terms. The mhctools license does not grant permission to use those materials. Check the predictor's linked upstream license before installing it or redistributing its code or weights.
The predictor matrix gives every predictor one of these:
| Tier | Meaning | What you do |
|---|---|---|
builtin |
Built into mhctools; nothing to download | Nothing. Calis and RandomBindingPredictor |
open |
Open source; code and weights can be fetched or installed freely | mhctools fetch <name> or pip install |
academic |
Academic / non-commercial terms, often with no redistribution | Read the upstream license, then mhctools fetch <name> --accept-license or install it yourself |
dtu |
DTU academic license, bound to an identity | Request it from DTU; mhctools calls your installation |
unlicensed |
Upstream publishes no license | --accept-license records that you have confirmed your own use is authorized |
What --accept-license means¶
It records that you reviewed the terms and confirmed that your own use is authorized. It does not grant rights mhctools does not have, and it cannot stand in for a license you must request yourself.
The DTU downloads (NetMHCpan, NetMHC, NetMHCcons, NetMHCIIpan, NetMHCstabpan,
NetChop) are the clearest example. DTU requires a name, position, academic
email, affiliation and acceptance, then sends a private link, so those
installations stay manual in mhctools ls and fetch will not install them.
NetCleave and TLimmuno2 publish no license at all. mhctools can fetch a
pinned snapshot, but only after --accept-license; the recorded manifest says
"license": "none published".
Copyleft tools¶
ERAMER, Tulip and PlifePred2 are GPLv3. mhctools vendors none of them and
does not import them into its own interpreter: ERAMER's workbook is read at
runtime, TULIP runs out of process through its own predict.py, and
PlifePred2 and Pfeature run in a separate interpreter.
Ambiguous upstream licenses¶
PeptiVerse declares Apache-2.0 on its model card and MIT in its README;
mhctools lists it as open but you should confirm which applies to your use.
Per-predictor details¶
| Topic | Where |
|---|---|
| MixMHCpred, MixMHC2pred, PRIME, MixTCRpred terms | their sections in binding, immunogenicity and TCR |
| SMM bundle (Non-Profit Open Software License 3.0) | SMM setup |
How fetch records provenance |
getting models |