Sequence -----------+--> 3D_structure --> 3D_module --+ +--> ?
| | | +--> ?
| | +--> Joint_module --> Hierarchical_CLF +--> ?
| | | +--> ?
+-> NLP_embeddings -+-------> Embedding_module -------+ +--> ?
ClynMut: Predicting the Clynical Relevance of Genome Mutations (wip)
To be a next-generation DL-based phenotype prediction from genome mutations. Will use sota NLP and structural techniques.
Planned modules will likely be:
- 3D learning module
- NLP embeddings
- Joint module + Hierarchical classification
The main idea is for the model to learn the prediction in an end-to-end fashion.
Install
$ pip install clynmut
Example Usage:
import torch
from clynmut import *
hier_graph = {"class": "all",
"children": [
{"class": "effect_1", "children": [
{"class": "effect_12", "children": []},
{"class": "effect_13", "children": []}
]},
{"class": "effect_2", "children": []},
{"class": "effect_3", "children": []},
]}
model = MutPredict(
seq_embedd_dim = 512,
struct_embedd_dim = 256,
seq_reason_dim = 512,
struct_reason_dim = 256,
hier_graph = hier_graph,
dropout = 0.0,
use_msa = False,
device = None)
seqs = ["AFTQRWHDLKEIMNIDALTWER",
"GHITSMNWILWVYGFLE"]
pred_dicts = model(seqs, pred_format="dict")
Important topics:
3D structure learning
There are a couple architectures that can be used here. I've been working on 2 of them, which are likely to be used here:
Hierarchical classification
- A simple custom helper class has been developed for it.
Testing
$ python setup.py test
Datasets:
This package will use the awesome work by Jonathan King at this repository.
To install
$ pip install git+https://github.com/jonathanking/sidechainnet.git
Or
$ git clone https://github.com/jonathanking/sidechainnet.git
$ cd sidechainnet && pip install -e .
Citations:
@article{pejaver_urresti_lugo-martinez_pagel_lin_nam_mort_cooper_sebat_iakoucheva et al._2020,
title={Inferring the molecular and phenotypic impact of amino acid variants with MutPred2},
volume={11},
DOI={10.1038/s41467-020-19669-x},
number={1},
journal={Nature Communications},
author={Pejaver, Vikas and Urresti, Jorge and Lugo-Martinez, Jose and Pagel, Kymberleigh A. and Lin, Guan Ning and Nam, Hyun-Jun and Mort, Matthew and Cooper, David N. and Sebat, Jonathan and Iakoucheva, Lilia M. et al.},
year={2020}
@article{rehmat_farooq_kumar_ul hussain_naveed_2020,
title={Predicting the pathogenicity of protein coding mutations using Natural Language Processing},
DOI={10.1109/embc44109.2020.9175781},
journal={2020 42nd Annual International Conference of the IEEE Engineering in Medicine & Biology Society (EMBC)},
author={Rehmat, Naeem and Farooq, Hammad and Kumar, Sanjay and ul Hussain, Sibt and Naveed, Hammad},
year={2020}
@article{pagel_antaki_lian_mort_cooper_sebat_iakoucheva_mooney_radivojac_2019,
title={Pathogenicity and functional impact of non-frameshifting insertion/deletion variation in the human genome},
volume={15},
DOI={10.1371/journal.pcbi.1007112},
number={6},
journal={PLOS Computational Biology},
author={Pagel, Kymberleigh A. and Antaki, Danny and Lian, AoJie and Mort, Matthew and Cooper, David N. and Sebat, Jonathan and Iakoucheva, Lilia M. and Mooney, Sean D. and Radivojac, Predrag},
year={2019},
pages={e1007112}
409 Oct 28, 2022
211 Dec 28, 2022
83 Jan 9, 2023
2.1k Feb 17, 2021
138 Dec 30, 2022
286 Dec 6, 2022
6.5k Jan 8, 2023
17 Dec 23, 2022
2 Jun 10, 2022
20 Nov 30, 2022
488 Nov 28, 2022
678 Jan 05, 2023
10 Jul 01, 2022
132 Nov 14, 2022
0 Oct 25, 2022
1 Jun 05, 2022
9 Nov 09, 2021
94 Dec 30, 2022
3 Apr 15, 2022
74 Oct 07, 2022
79 Nov 29, 2022
1 Oct 22, 2021
3 Apr 02, 2022
49 Dec 06, 2022
23 May 19, 2022