Phosphoproteome Prediction

The software tool Phosphoproteome Prediction is a computational method that predicts phosphorylation levels of proteins across cancer patients using available proteomic, transcriptomic, and genomic data. The tool integrates four key components:

1. Baseline correlations between protein and phosphoprotein abundances

2. Universal protein-protein interactions

3. Shareable regulatory information across cancer tissues

4. Associations among multi-phosphorylation sites of the same protein

The method was developed as part of the 2017 NCI-CPTAC DREAM Proteogenomics Challenge and ranked first in predicting phosphorylation levels in both breast and ovarian cancer samples, demonstrating its robustness and generalization ability. The tool serves as an alternative to the time-consuming and expensive mass spectrometry-based phosphoproteomics technique, which requires specialized expertise and a large amount of starting material.

Topic

Oncology;Proteomics;Machine learning

Detail

  • Operation: Post-translation modification site prediction;iTRAQ;Protein interaction prediction

  • Software interface: Command-line user interface

  • Language: R,Shell,Python

  • License: Not stated

  • Cost: Free of charge

  • Version name: -

  • Credit: NSF, American Heart Association, Amazon Web Services3.0 Data Grant Portfolio: Artificial Intelligence and Machine Learning Training Grants.

  • Input: -

  • Output: -

  • Contact: Hongyang Li hyangl@umich.edu, Yuanfang Guan gyuanfan@umich.edu

  • Collection: -

  • Maturity: -

Publications

  • Machine learning empowers phosphoproteome prediction in cancers.
  • Li H and Guan Y. Machine learning empowers phosphoproteome prediction in cancers. Machine learning empowers phosphoproteome prediction in cancers. 2020; 36:859-864. doi: 10.1093/bioinformatics/btz639
  • https://doi.org/10.1093/BIOINFORMATICS/BTZ639
  • PMID: 31410451
  • PMC: PMC7868059

Download and documentation

< Back to DB search