Evaluating Mouse Gene Function Predictions
Author Information
Author(s): Peña-Castillo Lourdes, Tasan Murat, Myers Chad L, Lee Hyunju, Joshi Trupti, Zhang Chao, Guan Yuanfang, Leone Michele, Pagnani Andrea, Kim Wan Kyu, Krumpelman Chase, Tian Weidong, Obozinski Guillaume, Qi Yanjun, Mostafavi Sara, Lin Guan Ning, Berriz Gabriel F, Gibbons Francis D, Lanckriet Gert, Qiu Jian, Grant Charles, Barutcuoglu Zafer, Hill David P, Warde-Farley David, Grouios Chris, Ray Debajyoti, Blake Judith A, Deng Minghua, Jordan Michael I, Noble William S, Morris Quaid, Klein-Seetharaman Judith, Bar-Joseph Ziv, Chen Ting, Sun Fengzhu, Troyanskaya Olga G, Marcotte Edward M, Xu Dong, Hughes Timothy R, Roth Frederick P
Primary Institution: University of Toronto
Hypothesis
Can computational predictions of gene function improve our understanding of mouse genes?
Conclusion
The study shows that current data allows for accurate predictions of gene functions in mice, including many previously uncharacterized genes.
Supporting Evidence
- 76% of mouse genes had inferred functions.
- 5,000 previously uncharacterized genes were predicted to have functions.
- 41% average precision at a recall rate of 20%.
- 26% of GO terms achieved a precision better than 90%.
Takeaway
Scientists used computers to guess what mouse genes do, and they found out a lot of new things about them!
Methodology
Nine bioinformatics teams used a standardized dataset of mouse genomic data to train classifiers and predict gene functions based on Gene Ontology terms.
Potential Biases
Potential overfitting due to the limited number of training examples for some GO terms.
Limitations
The predictions may not be accurate for all genes, especially those with limited data.
Statistical Information
P-Value
p<0.05
Statistical Significance
p<0.05
Digital Object Identifier (DOI)
Want to read the original?
Access the complete publication on the publisher's website