NOTICE
======

The code in this repository is original work, licensed MIT (see LICENSE). The *data* is not
original: it is FoldX output computed on structures and mutations defined by SKEMPI 2.0. Those
have their own terms, and MIT does not and cannot override them.

LICENSE therefore covers the CODE. The computed results under skempi_foldx/data/ are derived
from SKEMPI 2.0 (CC BY 4.0) and were produced with FoldX, which is separately licensed. Read
this file before redistributing them.


Data provenance
---------------

  * **SKEMPI 2.0** supplies every mutation, every chain grouping, and the structures the pipeline
    runs on. Nothing here would exist without it.

        Jankauskaite, J., Jimenez-Garcia, B., Dapkunas, J., Fernandez-Recio, J. & Moal, I. H.
        "SKEMPI 2.0: an updated benchmark of changes in protein-protein binding energy,
        kinetics and thermodynamics upon mutation."
        Bioinformatics 35(3):462-469, 2019.  doi:10.1093/bioinformatics/bty635

        https://life.bsc.es/pid/skempi2

    SKEMPI 2.0 is released under **CC BY 4.0**
    (https://creativecommons.org/licenses/by/4.0/). What is redistributed here is an *adaptation*:
    `skempi_foldx/data/results_{sp,mp}/` carries SKEMPI's per-complex mutation lists, alongside the
    energies computed here. Every record carries SKEMPI's verbatim `cleaned` mutation string as a
    field, and every record key is that same string; a record whose role-chain name differs from
    it also carries that name in a `role` field, derived from SKEMPI's chain mapping files rather
    than copied from them. What is **not** redistributed: SKEMPI's measured affinities,
    its structures, or the database itself, all of which come from BSC and are required to run the
    pipeline. CC BY 4.0 permits the adaptation and requires, under §3(a)(1), that the creator be
    identified, that a URI to the licensed material be given, that the licence be identified, and
    that the modification be indicated. The citation and link above supply the first two; this
    paragraph supplies the last two.

  * **Protein Data Bank** structures, via SKEMPI's cleaned PDB set. Not redistributed here.

  * **FoldX 5.1** produced every number in `skempi_foldx/data/`. FoldX is licensed software from
    the CRG -- free to academic and non-profit research institutions, paid for commercial use --
    (https://foldxsuite.crg.eu/) and is **not** included, wrapped, or redistributed in any form.
    Running the pipeline requires a separately obtained licence and binary.


A caveat on the computed energies
---------------------------------

The MIT licence in LICENSE covers this repository's code. Whether FoldX's own licence permits
redistribution of energies computed with it is a question for the CRG's terms, not something this
file can settle. The values are shipped here on the reading that they are derived measurements
rather than a redistribution of the software, and because the alternative -- every group
re-purchasing a licence and re-spending the CPU-weeks to recompute the same deterministic numbers
-- serves nobody. Relicensing or redistributing this data commercially requires checking the CRG
terms directly rather than relying on this paragraph.


Full citations with DOIs for every work named in this file are collected in docs/REFERENCES.md,
together with the data sources, tools and comparator methods named under docs/. That file is in
the source tree and the sdist but not the wheel, so from an installed copy read it at
https://github.com/cchin29/skempi-foldx/blob/v0.2.2/docs/REFERENCES.md


Related work
------------

  * This package was extracted from a research fork of **MuLAN** (Lombardi & Carbone), where it
    began as a score channel for a ddG predictor. Upstream MuLAN is licensed CC BY-NC-SA 4.0, and
    the fork inherited that; **none of its code is present here**, which is what leaves this
    repository free to carry a permissive licence. The extraction is deliberate: the FoldX
    pipeline and its results are useful independently of any one model, and a non-commercial
    licence inherited from the parent project would have prevented that.

        Lombardi, G. & Carbone, A. "MuLAN: Mutation-driven Light Attention Networks for
        investigating protein-protein interactions from sequences."
        bioRxiv 2024.08.24.609515 (2024).  doi:10.1101/2024.08.24.609515

        https://github.com/GianLMB/mulan

  * The mutation exclusion for `1KBH` follows **RDE-Network**'s `block_list`; see
    `skempi_foldx/exclusions.py` for the evidence and the reasoning.

        Luo, S., Su, Y., Wu, Z., Su, C., Peng, J. & Ma, J. "Rotamer Density Estimator is an
        Unsupervised Learner of the Effect of Mutations on Protein-Protein Interaction."
        ICLR 2023; bioRxiv 2023.02.28.530137.  doi:10.1101/2023.02.28.530137

        https://github.com/luost26/RDE-PPI
