Using Machine Learning Approaches to Predict Target Gene Expression in Rice T-DNA Insertional Mutants.

Area of Science:

• Plant genomics
• Computational biology
• Functional genomics

Background:

• T-DNA insertion is a common method for rice functional gene research.
• Experimental validation is required to confirm gene expression changes in T-DNA mutants.
• Improving screening efficiency for activated genes is crucial.

Purpose of the Study:

• To establish a machine learning model for predicting gene expression in T-DNA mutants.
• To enhance the efficiency of screening for activated genes in rice.

Main Methods:

• Gathered experimental datasets of gene expression in T-DNA mutants.
• Captured PROMOTER and MIDDLE sequences for encoding.
• Developed a two-layer machine learning model (TIMgo) using SVM, logistic regression, mRMR, and LADTree algorithms.

Main Results:

• The TIMgo model achieved 99.3% accuracy via cross-validation and 85.6% on independent testing.
• Local sequence information was more critical than global sequence information for classification.
• TIMgo demonstrated predictive ability for genes within 20 kb of the 35S enhancer.
• Identified the G-box regulatory sequence as potentially important for 35S enhancer activation.

Conclusions:

• The TIMgo model significantly improves the prediction of gene expression in T-DNA mutants.
• Local sequence features are key predictors of gene expression changes.
• The findings provide insights into T-DNA insertion effects and enhancer mechanisms in rice.