Extending Biological Pathways by Utilizing Conditional Mutual Information Extracted from RNA-SEQ Gene Expression Data

Multi-omics approaches for biosynthetic pathway prediction in plants

“…One approach to overcome this obstacle is using coexpression networks; they are one of the multiple types of gene networks that can be generated from gene expression data 39,40 , and represent genes as nodes connected by edges/links if the genes show coexpression correlation above a predetermined threshold. In A. thaliana, coexpression networks have been shown to capture the functional categorization of genes by grouping them in clusters or modules of tightly coexpressed genes, which have similar function or regulation 41 .…”

“…Moreover, evidence suggest that the function and members of some gene modules are conserved through speciation events 42 , allowing the propagation of annotations between species with more confidence. When handling large datasets, however, coexpression networks can become too complex to readily interpret 22,39,43 , in part due to unobserved factors that affect gene expression and may cause correlation among genes 44 , which in turn may lead to an increase in type I errors when interpreting the significance of correlated genes in a network or module. In summary, better gene prioritization methods need to be developed to improve the confidence of predictions from the analysis of coexpression networks.…”

“…These outliers all corresponded to recent BGCs were found in dicot genomes. These outliers all corresponded to recent (partial) genome amplification events, such as in the case of Camelina sativa (37) with 88 candidate BGCs, Brassica napus (38) with 68 candidate BGCs and G. max (39) with 76 candidate BGCs. In many plant genomes, candidate BGCs of high complexity were identified, with as many as seven or eight different enzymatic classes encoded in the same tight genomic region.…”