sarscov2vec

Realize Elton et al. pipeline using Mekni et al. SARS-CoV-2 viral protease SVM on PubMed Central PMC Open Access articles.

scheme

project

ChemDataExtractor is used to identify Chemical Entities validate using PubChemPy and PaDEL-Descriptor software to extract compunds descriptors.

description

2-d PCA is used to plot word2vec results following Elton et al. pipeline. Moreover, as different approach, elbow method to select optimal out PCA dimension is followed and incremental K-means is applied.

design

Strategy pattern is followed to dynamically change behavior on different load/store strategies and classifiers.

usage

foo@bar:~/project$ ./build.sh
...
# start padel container
foo@bar:~/project$ ./padel-service/padel-service.sh 
...
# start mongo docker container
foo@bar:~/project$ ./mongo-dock.sh
...
# start project
foo@bar:~/project$ python3 sarscov2vec.py
...

Optionally is possible to remove lines in code/mainProject.py (commented with "delete this to use FS") to disable usage of MongoDB and use file system to store chemical entities and sentences. In this case skip start mongo docker container command.

results

pca 2-d

PCA 2-d results coloring active compunds against SARS-CoV-2 viral protease.

optimal PCA out and K-MEANS

Cluster Num.	active CE	CE	words
0	15	989	89879
1	1	92	4370
2	0	9	1272

Coeff.	value
silhouette avg	0.8479288100264025
SSE (k=3)	2390

Term	t0	t1	t2	t3	t4	t5	t6	t7	t8	t9
covid-19	ill	psychiatric	hiv-positive	dementia	pandemic	concern	pertain	hemophilia	people	behaviour

identified active fragments

MarcoGarlet/sarscov2vec