PureseqTM

TransMembrane (TM) topology prediction from amino acid sequence only

Server

Users may submit their sequences to the PureseqTM Server at: http://pureseqtm.predmp.com

Train and test data

Users may find the train, test data of PureseqTM at: https://github.com/PureseqTM/PureseqTM_Dataset

Paper

https://www.biorxiv.org/content/10.1101/627307v2.abstract

Contact email

pureseqtm@gmail.com

Install

Download the PureseqTM package

git clone https://github.com/PureseqTM/PureseqTM_Package.git
cd ./PureseqTM_Package

Compilation if necessary

cd source_code
    make
cd ../

Examples

single protein prediction

prediction with fast mode (i.e., NOT use DBN features from Philius)

./PureseqTM.sh -i example/4j7cK.fasta -m 0

prediction without ground-truth label

./PureseqTM.sh -i example/4j7cK.fasta

given ground-truth 2-state TM label

./PureseqTM.sh -i example/4j7cK.fasta -l example/4j7cK.top

evaluate the prediction result

If the ground-truth label is provided, run below commands to evaluate the prediction accuracy:

grep -v "#" 4j7cK_PureTM/4j7cK.prob | awk '{print $NF" "$3}' > 4j7cK.pred_reso
util/TM2_Evaluation 4j7cK.pred_reso 0.5
rm -f 4j7cK.pred_reso

plot the posterior probabilities

./PureseqTM.sh -i example/4j7cK.fasta -P 1

Note that gnuplot is required to be installed in the local system.

prediction with given profile

./PureseqTM.sh -i example/1bhaA.tgt

Note that to generate the evolutionary profile, users are suggested to use the TGT_Package to generate the TGT file:

./A3M_TGT_Gen.sh -i <input_fasta> -d uniprot20_2016_02 -h hhsuite2 -n 3 -E 10

For more details on evolutionary profile, please goto https://github.com/realbigws/TGT_Package

proteome prediction

As it takes about 2 hours for the whole Human proteome, below please find a toy example:

./PureseqTM_proteome.sh -i example/test_proteome.fasta

transfer PDBTM label

Transfer PDBTM 9-state label to 0/1 TransMembrane (TM) label:

python util/pdbtm2binary.py example/1bhaA.pdbtm

Output files

There shall be 5 to 7 output files in XXX_PureTM by default, where XXX is the input name:

File name	Description	Option
XXX.fasta_raw	original input sequence file in FASTA format.
XXX.fasta	canonical sequence file without non-canonical characters.
XXX.top	simple 2-state TransMembrane (TM) prediction in FASTA format.
XXX.prob	detailed 2-state TransMembrane (TM) probability prediction.
XXX.pred_mode	prediction mode (1 for sequence and 0 for profile).
XXX.png	posterior probabilities plotted by GNUPLOT	if option -P 1 is set
XXX.gff	segment-level output	if option -P 1 is set

References

Phobius

Title:
     A Combined Transmembrane Topology and Signal Peptide Prediction Method
Authors:
     Lukas Kall, Anders Krogh and Erik L. L. Sonnhammer
Journal:
     J. Mol. Biol. (2004) 338, 1027–1036

Philius

Title:
     Transmembrane Topology and Signal Peptide Prediction Using Dynamic Bayesian Networks
Authors:
     Sheila M. Reynolds, Lukas Kall, Michael E. Riffle, Jeff A. Bilmes, William Stafford Noble
Journal:
     PLoS computational biology (2008) 4(11):e1000213

DeepCNF

Title:
     Protein secondary structure prediction using deep convolutional neural fields
Authors:
     Sheng Wang, Jian Peng, Jianzhu Ma, Jinbo Xu
Journal:
     Scientific reports (2016) 6:18962