DR MUHAMMAD SHAFIQ

• Machine Learning + Genome-Wide Analysis
• Predicting Gene Family Functions
• (In Silico Approach)

WHY PREDICT GENE FUNCTION?

• • Wet-lab experiments are costly & time-consuming
• • Genome sequencing is faster than functional validation
• • Need computational prediction before experiments

WHAT IS GENOME-WIDE ANALYSIS?

• • Identification of all gene family members
• • Domain & motif analysis
• • Phylogenetic relationships
• • Gene structure & duplication

DATA USED FOR ML MODELS

• • DNA & protein sequences
• • Conserved domains (Pfam/SMART)
• • RNA-Seq expression profiles
• • Promoter cis-elements
• • Subcellular localization

HOW MACHINE LEARNING HELPS

• • Converts gene features into numerical data
• • Learns patterns from known genes
• • Predicts functions of unknown genes

POPULAR ML ALGORITHMS

• • Random Forest
• • Support Vector Machine (SVM)
• • Neural Networks
• • Gradient Boosting
• • Graph-based learning

CO-EXPRESSION NETWORK APPROACH

• • Genes with similar expression → similar function
• • Network clustering + ML
• • Guilt-by-association principle

CROSS-SPECIES PREDICTION

• • Models trained on Arabidopsis/Rice
• • Applied to non-model plants
• • Conserved gene families

WHY WET-LAB NOT NEEDED INITIALLY

• • In silico predictions reduce experiments
• • Rank top candidate genes
• • Focus validation on most important genes

APPLICATIONS

• • Stress-responsive gene families
• • Metabolic pathways
• • Transcription factor families
• • Crop improvement

CONCLUSION

• • Genome-wide analysis provides data
• • Machine learning finds hidden patterns
• • Together they predict gene functions efficiently