PCTpred

PCTpred (PTM Cross-Talk predictor) is a computational method that can accurately identify PTM cross-talk pairs in a given protein sequence or structure. To this end, we first design a group of novel residue pair- and residue-based features which effectively show the preferences of cross-talk sites from both the sequence and structural perspectives. Then we construct two component predictors using random forest with finely selected sequence- and structure-based features. Further combining these two predictors, PCTpred takes full advantage of the complementarity between residue pair- and residue-based features and that between sequence and structural information. Using both pair- and protein-based evaluations, PCTpred yields consistently better results than the state-of-the-art methods.

Installation

Dependencies

• Perl ( >= 5.0 ) and file::copy, statistics::descriptive, pdl, bioperl, moose modules
• Python ( >= 2.7 ) and scikit-learn, numpy, networkx, biopython modules
• Matlab ( >= R2014a )
• GCC ( >= 5.3.0 )

Download

git clone https://github.com/Liulab-HZAU/PCTpred.git

User installation

To install PCTpred, you need to conduct the installation of some third-party software, including BLAST+, EVfold, DisEMBL, HMMER, SIFT, PolyPhen-2, Fpocket2, DSSP, PSAIA, and HBPLUS.

1. You need to install BLAST+ and add it into the environment variables and download non-redundant database for BLAST searches.

2. You can run the following shell script to install EVfold, DisEMBL, HMMER, Fpocket2, DSSP, and PSAIA.

sh install_software.sh

3. SIFT and PolyPhen-2 depend on several large databases, and you need to install them by yourself.

cd ./software/sift/sift6.2.1/  to install SIFT

cd ./software/polyphen-2/polyphen-2.2.2/  to install PolyPhen-2

4. HBPLUS is freely available for academic use. You need to complete the confidentiality agreement and send an email to receive the download instructions. You can install hbplus in ./software/hbplus/. If you want to skip this step, you may need to modify the path in PCTpred mentioned later.

How to use PCTpred ?

Check the software and databases path

• You will see a PCTpred.pl script in the ./bin/ directory, and you need to check and modify the paths related to specific software and databases involved in this script.

Input file

• The format of the input file is as follows. You can also check the prefile.txt file in the examples directory. For proteins without structural information, the PDB_ID and PDB_chain information is denoted by NA.

P84243	T	12	K	10	2l43	A
P62805	K	21	K	17	5ja4	B
P37840	K	12	S	129	2n0a	B
P46937	S	397	S	400	NA	NA
P04637	S	215	S	392	NA	NA
O14920	S	177	K	163	NA	NA

Run it

• Go to ./bin/ directory, and run the following command to predict the probability score of samples being PTM cross-talk pairs.

  source environment_variable.sh    # add environment variable

  perl PCTpred.pl -f prediction_file -seq /the/directory/of/sequence/ -str /the/directory/of/PDB/ -out /the/directory/of/output/

  -f            the path of prediction file, such as ../examples/prefile.txt

  -seq       the path of sequence file (.fasta)

  -str         the path of structure file (.pdb)

  -out        the path of prediction result

• For example

  perl PCTpred.pl -f ../examples/prefile.txt -seq ../sequences/ -str ../PDB/ -out ../examples/

Prediction result

• You will obtain three output files in the directory /the/directory/of/output/, such as predicted_results.txt, seqfeature.txt, and strfeature.txt. The first file provides the prediction results and the second and third files provide the sequence features and structural features for each PTM pair generated by our method, respectively.
• predicted_results.txt contains the expected output as follows

P84243	T	12	K	10	2l43	A	prediction_score	0.977
P62805	K	21	K	17	5ja4	B	prediction_score	0.923
P37840	K	12	S	129	2n0a	B	prediction_score	0.841
P46937	S	397	S	400	NA	NA	prediction_score	0.992
P04637	S	215	S	392	NA	NA	prediction_score	0.827
O14920	S	177	K	163	NA	NA	prediction_score	0.867

Help and Support

Contact

• If you have any questions or suggestions about PCTpred, please contact us by email: [email protected] or [email protected]
• This software is free for academic use. For commercial use, please contact with the authors.

Citation

• Hui-Fang Liu and Rong Liu. Structure-based prediction of post-translational modification cross-talk within proteins using complementary residue- and residue pair-based features. Briefings in Bioinformatics. 2018, in press.