PTM-X is an algorithm to predict post-translational modification(PTM) cross-talk both intra- and inter- protein. In current version, the following four features are used to predict PTM cross-talk inter proteins, via a random forest classifier.
1. sequence residue co-evolution: The coevolution between two the amino acids across
multiple vertebrates. The multiple sequence alignment data is downloaded from
eggNOG v4.5. Here, the
1-Hamming loss
is used to measure the co-evolution score.
2. sequence motif co-evolution: The +/-3 surrounding amino acids are used to form
the motif for a given PTM locus. The fraction of consistence the this motif in
any vertebrate comparing to its human orthologous protein. Based on the two
vectors of motif conservation fraction, the co-conservation score is calculated
by the mean product of the two vectors. Again, the eggNOG v4.5 is used here.
3. Co-modification across different species: The co-conservation of PTM existence across human,
mouse and rat. The raw PTM data for these three species is downloaded from
PhosphoSitePlus.
4. Co-modification across different conditions: The co-occurrence between the two PTMs across 88
tissue, disease and cellline conditions.
Step 1: you need a Python environment for supporting packages. The easiest way might be installing the python platform via Anaconda. PTM-X is only compatible with Python 2.7. If you are using Anaconda 3, create a conda environment with Python 2.7 and dependent packages as follows,
conda create --name ptmxPy2 python=2.7 numpy scipy scikit-learn==0.17 joblib==0.11
# activate the environment
conda activate ptmxPy2Step 2: install PTM-X package by either downloading the source code here or use the following command line for an equivalent setting:
pip install -U git+https://github.com/huangyh09/PTM-XIn either case, if you don't have the root permission, add --user.
Step 3, to fetch above four featuresd ownload the data from http://ufpr.dl.sourceforge.net/project/ptm-crosstalk/PTM-X_data_v2.2.zip Unzip it and set the directory as the according parameter in the command lines below.
- To see all arguments, try command line
PTMX-feature -h - For the data_dir, set it as the downloaded data (see above).
- The python file is feature_extractor.py
- To see all arguments, try command line
PTMX-predict -h - The python file is predict.py.
- demo file: https://github.com/huangyh09/PTM-X/blob/master/demo.sh
- web server: http://bioinfo.bjmu.edu.cn/ptm-x/
- data repository: http://ptm-crosstalk.sourceforge.net
- For version 1 (intra-protein): Huang Y. et al.: Systematic Characterization and Prediction of Post-Translational Modification Cross-talk. Molecular & Cellular Proteomics, 2015, 14 (3), 761-770.
- For version 2 (inter-protein): Huang R. et al.: Systematic characterization and prediction of post-translational modification cross-talk between proteins. Bioinformatics, 2019, 35 (15), 2626-2633.