Gene Ontology Based Transfer Learning for Protein Subcellular Localization
Author Information
Author(s): Mei Suyu, Fei Wang, Zhou Shuigeng
Primary Institution: Shenyang Normal University, Shenyang, PR China; Shanghai Key Laboratory of Intelligent Information Processing, School of Computer Science, Fudan University, Shanghai, PR China
Hypothesis
Can a Gene Ontology Based Transfer Learning Model (GO-TLM) improve the prediction of protein subcellular localization?
Conclusion
The GO-TLM model significantly improves the accuracy of protein subcellular localization predictions compared to existing models.
Supporting Evidence
- GO-TLM achieved an accuracy of 96.65% on the MultiLoc plant dataset, significantly higher than the baseline model.
- Cross-validation results showed that GO-TLM outperformed MultiLoc-GO and Euk-mPLoc models.
- GO-TLM demonstrated stable performance across various subcellular locations.
Takeaway
This study created a new model to help scientists predict where proteins are located in a cell, which is important for understanding how they work.
Methodology
The study used a Gene Ontology Based Transfer Learning Model (GO-TLM) that integrates multiple data sources and employs non-parametric cross-validation for kernel weight estimation.
Potential Biases
Potential bias from the uneven distribution of GO terms and the reliance on evolutionary relationships.
Limitations
The model may introduce noise and outliers due to the reliance on homology-based GO term transfer.
Participant Demographics
The study evaluated proteins from various eukaryotic organisms, including plants, fungi, and animals.
Statistical Information
P-Value
12.98%
Statistical Significance
p<0.05
Digital Object Identifier (DOI)
Want to read the original?
Access the complete publication on the publisher's website