Modelling and Construction of Transcriptional Regulatory Networks using Time-series Gene Expression Data

Abstract

Selection of informative genes, modeling and construction of Transcriptional Regulatory Networks is an important problem in gene expression genomics. The small sample size and the large number of genes in gene expression data make the selection and modeling process complex. Further, the selected informative genes from high dimensional gene expression data may act as a vital input for genetic network analysis. The identification of hub genes and module interactions in genetic networks is yet to be fully explored. Usually, the raw gene expression data is taken as input for genetic network analysis, which is inherently noisy due to different sources of variation present in gene expression experiments. Further, these noises may mislead the results obtained from the network modeling and inference algorithms and techniques. Therefore, attempts are made to develop approaches for modeling and construction of Transcriptional Regulatory Networks after denoising the raw noisy gene expression data. In this study, a statistically sound gene selection technique based on support vector machine algorithm for selecting informative genes from high dimensional gene expression data was proposed. The comparative performance of the proposed gene selection technique (Boot-SVM-RFE) was evaluated on three different crop microarray datasets. The proposed gene selection technique outperformed most of the existing techniques for selecting robust set of informative genes. Further, the bootstrap procedure incorporated in this technique was able to remove the spurious association among genes and their corresponding classes. Here, attempts were also made to develop an algorithm to estimate the true gene expression value from raw expression matrix based on Wavelet methodology. Further, statistical approaches for modeling and construction of gene regulatory networks using vector autoregressive models and sparse autoregressive vector models were also developed using wavelet transformed gene expression data for time-series gene expression experiments. Further, the effect of levels (scales), filter types and filter lengths of various wavelet filters on gene regulatory network modeling and inferences was studied. For this purpose, extensive simulations (artificial gene expression data) and synthetic gene expression data (DREAM4 data for E. coli and S. cerviceae) were used. Though this, better combinations of wavelet decomposition levels, filter types and lengths for better modeling and inference of gene regulatory networks was obtained. Further, the comparative performance analysis of the proposed approach was carried out on DREAM4 data with respect to WGCNA, CLR, ARACNE, NetworkBMA and MVAR. The results indicated that our method performs better than these popular contemporary genetic network modeling and inference approaches. For identification hub genes in genetic networks is a crucial task in system biology. Therefore, an attempt has been made to develop a statistical approach for identification of hub genes in the gene co-expression network. Besides, a differential hub gene analysis approach has also been developed to group the identified hub genes into various groups based on their gene connectivity in a case vs. control study. Based on the proposed hub gene identification approach, a few number of hub genes were identified as compared to the existing approach, which is in accordance with the principle of scale free property of real networks. In this study, developed approaches were applied to salinity and aluminum stresses in rice and soybean respectively. Through this, various key genes revealed the underlying molecular mechanisms of salinity and Aluminum toxic stress response in rice and soybean were reported. Here, we developed two R packages, namely, dhga (https://cran.r-project.org/web/packages/dhga) and waveletGRN for the users.