A machine learning approach for genome-wide prediction of morbid and druggable human genes based on systems-level data
2010

Predicting Human Morbid and Druggable Genes Using Machine Learning

publication Evidence: moderate

Author Information

Author(s): Costa Pedro R, Acencio Marcio L, Lemke Ney

Primary Institution: Instituto de Biociências de Botucatu, UNESP - Univ Estadual Paulista

Hypothesis

Can a machine learning approach accurately predict morbid and druggable genes on a genome-wide scale?

Conclusion

The study demonstrates that network topological features, tissue expression profiles, and subcellular localization can reliably predict human morbid and druggable genes.

Supporting Evidence

  • The meta-classifier correctly recovered 65% of known morbid genes with a precision of 66%.
  • It correctly recovered 78% of known druggable genes with a precision of 75%.
  • Decision trees revealed that the number of regulating transcription factors is crucial for morbidity prediction.

Takeaway

Scientists used computers to find genes that can cause diseases or be targeted by drugs, making it easier to discover new treatments.

Methodology

A decision tree-based meta-classifier was trained on datasets containing network topological features, tissue expression profiles, and subcellular localization data.

Potential Biases

The lack of negative examples may lead to misclassification of unknown morbid genes as non-morbid.

Limitations

The approach relies on existing gene interaction data, which may not cover all human genes, and assumes that all genes not known to be morbid or druggable are non-morbid and non-druggable.

Statistical Information

P-Value

0.05

Statistical Significance

p<0.05

Digital Object Identifier (DOI)

10.1186/1471-2164-11-S5-S9

Want to read the original?

Access the complete publication on the publisher's website

View Original Publication