denovo-checkpoints / README.md
BioGeek's picture
Drop the What is deliberately NOT here section
8fcef57
|
Raw History Blame Contribute Delete
3.62 kB
---
license: other
license_name: see-per-file-provenance
pretty_name: De Novo Peptide Sequencing Checkpoint Mirror
tags:
- proteomics
- mass-spectrometry
- de-novo-peptide-sequencing
---
# De novo peptide sequencing: checkpoint mirror
A backup copy of published model checkpoints for *de novo* peptide sequencing,
kept because several of the originals live on Google Drive and Dropbox links
that carry no DOI, no version and no preservation commitment, and which
therefore can disappear without leaving anything to cite.
**These are other people's weights.** Nothing here was trained by this mirror.
Every file keeps the name its authors gave it, and the table below records where
it came from, which paper it belongs to and what licence its source repository
states. If you use a checkpoint, cite the paper, not this mirror.
Every known checkpoint for these methods, mirrored here or not, is listed with
its version, training data, licence and link state in
[**The weights behind the numbers**](https://jeroen.vangoey.be/awesome_de_novo_peptide_sequencing/#the-weights-behind-the-numbers),
part of the [Awesome De Novo Peptide Sequencing](https://jeroen.vangoey.be/awesome_de_novo_peptide_sequencing/) catalog. Each method's
own page there carries the same for that method alone.
## Provenance and licence, per file
| Folder | File | Method | Source | Licence of source repo |
| --- | --- | --- | --- | --- |
| `contranovo/` | `ControNovo.ckpt` | ContraNovo | [BEAM-Labs/denovo](https://github.com/BEAM-Labs/denovo) (Google Drive) | MIT |
| `pi-primenovo/` | `model_massive.ckpt` | π-PrimeNovo | [PHOENIXcenter/pi-PrimeNovo](https://github.com/PHOENIXcenter/pi-PrimeNovo) (Google Drive) | MIT |
| `pi-primenovo/` | `PTM_Phosphorylation.ckpt` | π-PrimeNovo (phospho) | [PHOENIXcenter/pi-PrimeNovo/pi-PrimeNovo-PTM](https://github.com/PHOENIXcenter/pi-PrimeNovo/tree/main/pi-PrimeNovo-PTM) (Google Drive) | MIT |
| `refinenovo/` | `model_massivekb.ckpt` | RefineNovo-30M | [BEAM-Labs/denovo](https://github.com/BEAM-Labs/denovo) (Google Drive) | MIT |
| `ranknovo-crossnovo/` | `model_0.660.ckpt` | RankNovo **and** CrossNovo | [BEAM-Labs/denovo](https://github.com/BEAM-Labs/denovo) (Google Drive) | MIT |
| `dianovo/` | `*.pt` | DIANovo | [hearthewind/dianovo](https://github.com/hearthewind/dianovo) (git-LFS) | Apache-2.0 |
| `deepnovo/` | `translate.ckpt-48600.*` + `knapsack.npy.zip` | DeepNovo | [nh2tran/DeepNovo](https://github.com/nh2tran/DeepNovo) (Google Drive, **now gated**) | non-commercial use (custom) |
| `deepnovo-dia/` | DeepNovo-DIA model | DeepNovo-DIA | [nh2tran/DeepNovo-DIA](https://github.com/nh2tran/DeepNovo-DIA) (Google Drive) | non-commercial use (custom) |
**The two DeepNovo entries carry custom non-commercial licences.** The files
say:
> DeepNovo is publicly available for non-commercial uses. Copyright (C) 2017.
> Authors. All rights reserved.
> Copyright (C) 2018. Authors. All rights reserved. DeepNovo-DIA is provided
> free of charge for academic and non-commercial use.
So use for academic and non-commercial purposes is granted. Redistribution is
not *expressly* granted, and this mirror is non-commercial, attributed, and
exists because the DeepNovo link is now gated behind a Google sign-in. If you
are an author of either and would rather these were not mirrored, open an issue
on the catalog repository and they will be removed.
**One file serves two methods.** `BEAM-Labs/denovo` lists the same Google Drive
id for RankNovo and for CrossNovo, so either they ship one checkpoint or the
README has an error; the file is stored once, under a folder naming both.