Fused Graphical Lasso Recovers Flowering Time Mutation Genes in Arabidopsis thaliana

Abstract

Conventional breeding approaches that focus on yield under highly favorable nutrient conditions have resulted in reduced genetic and trait diversity in crops. Under the growing threat from climate change, the mining of novel genes in more resilient varieties can help dramatically improve trait improvement efforts. In this work, we propose the use of the joint graphical lasso for discovering genes responsible for desired phenotypic traits. We prove its efficiency by using gene expression data for wild type and delayed flowering mutants for the model plant. Arabidopsis thaliana shows that it recovers the mutation causing genes LNK1 and LNK2. Some novel interactions of these genes were also predicted. Observing the network level changes between two phenotypes can also help develop meaningful biological hypotheses regarding the novel functions of these genes. Now that this data analysis strategy has been validated in a model plant, it can be extended to crop plants to help identify the key genes for beneficial traits for crop improvement.

Genetic architecture of inter-specific and -generic grass hybrids by network analysis on multi-omics data

“…In this case, ℓ 1 -penalization has been one of the main techniques used to make sparse inference in a Gaussian Markov random field (Friedmanet al 2007), yielding a sparse structured precision matrix ∑ −1 which, in turn, can be converted into an undirected network and further analyzed for its topological properties. This approach has been applied to the study of gene expression (Shahdoust et al 2019; Wu et al 2013) and metabolomic (Liu et al 2022) data in humans, with few examples in plants(Li and Jackson 2015; Kapoor et al 2021;e Lima et al 2018; Bartzis et al 2017). With a selected set of candidate features recovered from gene co-expression and metabolic networks, one can perform omic-phenotype integration.…”