Skip to content

Latest commit

 

History

History
57 lines (46 loc) · 1.57 KB

README.md

File metadata and controls

57 lines (46 loc) · 1.57 KB
Raw

Efflux ATP Binding

Usage

Preprocessing

Generate PSSM using Blast

  1. Make sure you've already the fasta files from Uniprot
  2. Make sure you've already set blast tool and NR DB
  3. run this command to generate pssm from fasta using blast
python pssmbyblast.py <fastadir>

Generate position list of binding site

python generateposlist.py

Filter pssm based on the similarity result

  1. Make sure you've already run blastclust tool to generate similarity result
  2. run this script to filter the pssm files
python filterpssm.py <resultsimilarityfile>

Generate Libsvm from pssm files

  1. run this script to generate libsvm files from pssm
python pssm_to_libsvm.py <folderofpssm> <windowssize>

Generate csv from libsvm

  1. run this script to generate csv as our final dataset
python libsvm_to_csv.py <folderoflibsvm>

Generate Dataset

  1. run this script to generate dataset as training, validation and independent with customize ratio
python generatedataset.py <folderofcsv> <ratio>

Run some algorithm

We provide some algorithm to try such as CNN, ResNet, VGG and DenseNet.

See this current result.

Algo ACC TN FN TP FP Specitivity Sensitivity MCC AUC
CNN 100 % 4177 1 48 1 100 % 98 % 0.979
ResNet 18 99 % 4176 1 48 2 100 % 98 % 0.969
ResNet 50 99 % 4176 1 48 2 100 % 98 % 0.969
VGG 16
VGG 19
DenseNet