PTM-X: A Web Server for PTM Cross-talk Prediction



Introduction introduction.png

Protein post-translational modifications (PTMs) add a further layer of complexity to the proteome and regulate a wide range of cellular protein functions. Many PTM sites from the same (intra) or different (inter) proteins are not isolated, and instead may coordinate with each other to determine a functional outcome, which is defined as PTM cross-talk.

PTM-X is a web server to predict PTM cross-talk both intra- and inter- protein.

Predict intra-protein predict.png

This service can predict intra- protein PTM crosstalk by inputting the Swiss-Prot accession number and the two candidate sites. The input data format was as the example shows:(columns are separated by Tab):

P04637	S15	phosphorylation	T18	phosphorylation

(1)The first column is the Swiss-Prot accession number.
(2)The second column is the position of first candidate site.
(3)The third column is the PTM type of the first position.
(4)The forth column is the position of second candidate site.
(5)The fifth column is the PTM type of the second position.
Note that unknown PTM type can be replaced by nan.

Users can use the template script provided for the prediction of intra-protein PTM cross-talk. More details in Download page.

Method

We collected 193 PTM cross-talk pairs in 77 human proteins from the literature, and then tested location preference, and co-evolution at the residue and modification levels. We found that cross-talk events preferentially occurred among nearby PTM sites, especially in disordered protein regions; and cross-talk pairs tended to co-evolve. Given the properties of intra- protein PTM cross-talk pairs, a naïve Bayes classifier integrating different features was built to predict cross-talks for pairwise combination of PTM sites intra protein.

Performance

During the performance evaluation, we found that the AUC( area of under the overall receiver operating characteristic (ROC) curve) in the cross-validation is 0.833. And AUC on an independent test set with 80 positive samples and 4169 negative samples achieves 0.74.

Predict inter-proteinpredict.png

This service can predict inter- proteins PTM crosstalk by inputting the pair of Swiss-Prot accession numbers and the corresponding candidate sites. The input data format was as the example shows: (columns are separated by Tab):

P12814	Y12	phosphorylation	Q05397	Y397	phosphorylation

(1)The first column is the Swiss-Prot accession number of the first protein.
(2)The second column is the candidate PTM site corresponding to the first protein.
(3)The third column is the PTM type of the first position.
(4)The forth column is the Swiss-Prot accession number of the second protein.
(5)The fifth column is the candidate PTM site corresponding to the second protein.
(6)The sixth column is the PTM type of the second position.
Note that unknown PTM type can be replaced by nan.

Users can use the template script provided for the prediction of inter-protein PTM cross-talk. More details in Download page.

Method

We collected 199 PTM cross-talk pairs in 82 pair of human proteins and then tested sequence residue co-evolution, sequence motif co-evolution, co-modification across species and different conditions. The results clearly indicates that inter-protein cross-talk PTM pairs have higher sequence co-evolution at both PTM residue and motif levels. Also, we found that cross-talk PTMs have higher co-modification across multiple species and 88 human tissues or conditions.We further applied a random forest classifier to predict PTM crosstalk between proteins by integrating three predictive features: residue and motif co-evolution, co-modification across different conditions. The co-modification across species is not included due to its limited prediction power, though it shows a good potential but probably requires more completed PTM sets.

Performance

The AUC (area under the receiver operating characteristic (ROC) curve) of the inter-protein prediction in the cross-validation test is 0.81 with integrating these three features.

Feedback feedback.png

All comments, suggestions, questions, and bug reports are welcome. For inquiries, please send an e-mail to Tingting Li, Ph.D., Peking University Health Science Center via litt@hsc.pku.edu.cn.

Citation