MTF_prediction

Here is the code used to perform the analysis of the paper. Inside the folders you can find the awk files to obtain the proper transmembrane domain. Also, tf_running_captum.py is the python modify version to obtain Captum interpretation.

UniProtKB 2022_05 Version Used

#MEMBRANE PROTEINS

(organism_id:9606) AND (reviewed:true) AND (keyword:KW-0472) #human -> 7663

(organism_id:10090) AND (reviewed:true) AND (keyword:KW-0472) #mouse -> 6527

FOR CYTOPLASMIC

Filter search from UniProt

(organism_id:9606) AND (reviewed:true) AND (ft_topo_dom:cytoplasmic) #human

(organism_id:10090) AND (reviewed:true) AND (ft_topo_dom:cytoplasmic) #mouse

#Download data GFF from UniProt

Filter to obtain only Cytoplasmic domains

grep "Cytoplasmic" cytoplasmic.gff > cytoplasmic_filtered.gff

AWK to adapt to UniProt format

awk -F "\t" '{print $1"["$4"-"$5"]"}' cytoplasmic_filtered.gff > cytoplasmic_filtered_IDs.csv

#Download fasta with these coords with "Retrieve/ID mapping", from "UniProtKB AC/ID" to "UniProtKB", format: FASTA (saurce list)

Run DeepTFactor

conda activate deeptfactor python tf_running.py -i ~/cytoplasmic.fasta -g cuda -o ~/result

Filter the positive results

grep "True" prediction_result.txt | sort -rn -k3 > cytoplasmic_true.csv

Extract UniProtKB AC/IDs to download GeneNames with "Retrieve/ID mapping"

awk '{print $1}' cytoplasmic_true.csv | cut -c 4- | cut -f1 -d"|" > uniprot_names.csv

---

Same but substract half of the TMD

awk -f half_cytoplasmic.awk cytoplasmic.gff > cytoplasmic_filtered.gff

awk -F "\t" '{print $1"["$4"-"$5"]"}' cytoplasmic_filtered.gff > cytoplasmic_filtered_IDs.csv

python tf_running.py -i ~/cytoplasmic.fasta -g cuda -o ~/result

grep "True" prediction_result.txt | sort -rn -k3 > all_cytoplasmic_true.csv

awk '{print $1}' all_cytoplasmic_true.csv | cut -c 4- | cut -f1 -d"|" > uniprot_names.csv

#Manually eliminate the duplicates keeping the highest score

---

Same but substract full TMD

awk -f full_cytoplasmic.awk cytoplasmic.gff > cytoplasmic_filtered.gff

awk -F "\t" '{print $1"["$4"-"$5"]"}' cytoplasmic_filtered.gff > cytoplasmic_filtered_IDs.csv

python tf_running.py -i ~cytoplasmic.fasta -g cuda -o ~/result

grep "True" prediction_result.txt | sort -rn -k3 > all_cytoplasmic_true.csv

awk '{print $1}' all_cytoplasmic_true.csv | cut -c 4- | cut -f1 -d"|" > uniprot_names.csv

#Manually eliminate the duplicates keeping the highest score

---

---

FOR EXTRACELLULAR

Filter search from UniProt

(organism_id:9606) AND (reviewed:true) AND (ft_topo_dom:extracellular) #human

(organism_id:10090) AND (reviewed:true) AND (ft_topo_dom:extracellular) #mouse

Filter to obtain only Extracellular domains

grep "Extracellular" extracellular.gff > extracellular_filtered.gff

AWK to adapt to UniProt format

awk -F "\t" '{print $1"["$4"-"$5"]"}' extracellular_filtered.gff > extracellular_filtered_IDs.csv

#Download fasta with these coords with "Retrieve/ID mapping", from "UniProtKB AC/ID" to "UniProtKB", format: FASTA (saurce list)

Run DeepTFactor

conda activate deeptfactor

python tf_running.py -i ~/extracellular.fasta -g cuda -o ~/result

Filter the positive results

grep "True" prediction_result.txt | sort -rn -k3 > extracellular_true.csv

Extract UniProtKB AC/IDs to download GeneNames with "Retrieve/ID mapping"

awk '{print $1}' extracellular_true.csv | cut -c 4- | cut -f1 -d"|" > uniprot_names.csv

---

Same but substract half of the TMD

awk -f half_extracellular.awk extracellular.gff > extracellular_filtered.gff

awk -F "\t" '{print $1"["$4"-"$5"]"}' extracellular_filtered.gff > extracellular_filtered_IDs.csv

python tf_running.py -i ~/extracellular.fasta -g cuda -o ~/result

grep "True" prediction_result.txt | sort -rn -k3 > extracellular_true.csv

awk '{print $1}' extracellular_true.csv | cut -c 4- | cut -f1 -d"|" > uniprot_names.csv

#Manually eliminate the duplicates keeping the highest score

---

Same but substract full TMD

awk -f full_extracellular.awk extracellular.gff > extracellular_filtered.gff

awk -F "\t" '{print $1"["$4"-"$5"]"}' extracellular_filtered.gff > extracellular_filtered_IDs.csv

python tf_running.py -i ~/extracellular.fasta -g cuda -o ~/result

cd /home/victor/deeptfactor/2022_analysis_final/extracellular/human_extracellular_full/result

grep "True" prediction_result.txt | sort -rn -k3 > extracellular_true.csv

awk '{print $1}' extracellular_true.csv | cut -c 4- | cut -f1 -d"|" > uniprot_names.csv

#Manually eliminate the duplicates keeping the highest score

---

---

FOR LUMENAL

Filter search from UniProt

(organism_id:9606) AND (reviewed:true) AND (ft_topo_dom:lumenal) #human

(organism_id:10090) AND (reviewed:true) AND (ft_topo_dom:lumenal) #mouse

Filter to obtain only Lumenal domains

grep "Lumenal" lumenal.gff > lumenal_filtered.gff

AWK to adapt to UniProt format

awk -F "\t" '{print $1"["$4"-"$5"]"}' lumenal_filtered.gff > lumenal_filtered_IDs.csv

#Download fasta with these coords with "Retrieve/ID mapping", from "UniProtKB AC/ID" to "UniProtKB", format: FASTA (saurce list)

Run DeepTFactor

conda activate deeptfactor

python tf_running.py -i ~/lumenal.fasta -g cuda -o ~/result

Filter the positive results

grep "True" prediction_result.txt | sort -rn -k3 > lumenal_true.csv

Extract UniProtKB AC/IDs to download GeneNames with "Retrieve/ID mapping"

awk '{print $1}' lumenal_true.csv | cut -c 4- | cut -f1 -d"|" > uniprot_names.csv

---

Same but substract half of the TMD

awk -f half_lumenal.awk lumenal.gff > lumenal_filtered.gff

awk -F "\t" '{print $1"["$4"-"$5"]"}' lumenal_filtered.gff > lumenal_filtered_IDs.csv

python tf_running.py -i ~/lumenal.fasta -g cuda -o ~/result

grep "True" prediction_result.txt | sort -rn -k3 > lumenal_true.csv

awk '{print $1}' lumenal_true.csv | cut -c 4- | cut -f1 -d"|" > uniprot_names.csv

---

Same but substract full TMD

awk -f full_lumenal.awk lumenal.gff > lumenal_filtered.gff

awk -F "\t" '{print $1"["$4"-"$5"]"}' lumenal_filtered.gff > lumenal_filtered_IDs.csv

python tf_running.py -i ~/lumenal.fasta -g cuda -o ~/result

grep "True" prediction_result.txt | sort -rn -k3 > lumenal_true.csv

awk '{print $1}' lumenal_true.csv | cut -c 4- | cut -f1 -d"|" > uniprot_names.csv

#Manually eliminate the duplicates keeping the highest score

#Domains greater than 1000 amino acids were not analyzed due to the limitation of the tool.

---

---

Check errors (Proteins with "U" aas)

grep -hv "^>" extracellular.fasta > error.fasta

grep "U" error.fasta

Use of Integrated Gradients

python tf_running_captum.py -i ATF6A.fasta -g cuda -o ./results_captum

---

---

Citation

Please be advised that if you choose to utilize this software for any academic publication or other purposes, it is strongly recommended that you provide appropriate attribution.

Shield:

This work is licensed under a Creative Commons Attribution 4.0 International License.