BSc-MSc (Integrated)

Developing regulatory network of genes controlling traits which are of importance economically and commercially are gaining much significance in present times. GRN’s provide an insight into the transcriptional mechanisms that regulate the robust and stochastic gene expression and their relationship with the phenotypic variability that can be utilized for better crop improvement strategies. The former approaches for Gene Regulatory Network construction mainly rely on using gene expression data as input, but the time consumption and high cost of expression analysis paved way for developing new methodologies that make GRN prediction easier.
The integration of genomic information along with gene expression data, could make the process of Gene Regulatory Network (GRN) construction more reliable than using expression data alone as input source. Using this approach, we have tried to develop the regulatory network of genes controlling immunity in cassava with special context to Bacterial blight resistance. Initially the immunity related genes in cassava were identified by protein domain search and analysis using HMMER. Cassava specific genes were further filtered for high competency, mapped and annotated to determine its biological role and function. A set of 1919 immunity related genes in cassava were identified, out of which 22 of them were specifically conferring virus resistance, 727 of them were screened for bacterial blight resistance by microarray data integration and a network was created using they predicted interactions identified from 324 genes using STRING. The networks obtained was visualized using Cytoscape and cross validated with simulated dataset generated from SynTReN. The generated network if immunity related genes in cassava could give more insight into the defence mechanism in cassava that can help in adapting better crop improvement and management strategies.
A comparison of various approaches used for GRN prediction like probabilistic method, mutual information-based method, correlation-based approaches etc was also done and various tool like ARACNE, WGCNA etc were evaluated. Networks with different sizes, 50, 100 and 150 was generated and the network parameters like clustering coefficient, network density etc were compared. Clustering coefficient does not seem to vary with increase in network size but network heterogeneity and density were observed to increase. The statistical analysis of the performance of different methods resulted into a conclusion that the mutual information based approaches are better tools for Gene Regulatory Network construction than the other methods and it performed with a specificity of 75.7% and a sensitivity of 79.4%.