COMPUTATIONAL APPROACHES IN THE ESTIMATION AND ANALYSIS OF TRANSCRIPTS DIFFERENTIAL EXPRESSION AND SPLICING: APPLICATION TO SPINAL MUSCULAR ATROPHY

Spinal Muscular Atrophy (SMA) is among the most common genetic neurological diseases that cause infant mortality. SMA is caused by deletion or mutations in the survival motor neuron 1 gene (SMN1), which are expected to generate alterations in RNA transcription, or splicing and most importantly reductions in mRNA transport within the axons of motor neurons (MNs). SMA ultimately results in the selective degeneration of MNs in spinal cord, but the underlying reason is still not clear entirely. The aim of this study is to investigate splicing abnormalities in SMA, and to identify genes presenting differential splicing possibly involved in the pathogenesis of SMA at genome-wide level. We performed RNA-Sequencing data analysis on 2 SMA patients and 2 controls, with 2 biological replicates each sample, derived from their induced Pluripotent Stem Cells-differentiated-MNs. Three types of analyses were executed. Firstly, differential expression analysis was performed to identify possibly mis-regulated genes using Cufflinks. Secondly, alternative splicing analysis was conducted to find differentially-used exons (DUEs; using DEXSeq) as splicing patterns are known to be altered in MNs by the suboptimal levels of SMN protein. Thirdly, we did RNA-binding protein (RBP) - motif discovery for the set of identified alternative cassette-DUEs, to pinpoint possible mechanisms of such alterations, specific to MNs. The gene ontology enrichment analysis of significant DEGs and alternative cassette-DUEs revealed various interesting terms including axon-guidance, muscle-contraction, microtubule-based transport, axon-cargo transport, synapse etc. which suggests their involvement in SMA. Further, promising results were obtained from motif analysis which has identified 22 RBPs out of which 7 RBPs namely, PABPC1, PABPC3, PABPC4, PABPC5, PABPN1, SART3 and KHDRBS1 are known for mRNAs stabilization and mRNA transport across MN-axon. Five RBPs from PABP family are known to interact directly with SMN protein that enhance mRNA transport in MNs. To validate our results specific wet-lab experiments are required, involving precise recognition of RNA-binding sites correspondent with our findings. Our work has provided a promising set of putative targets which might offer potential therapeutic role towards treating SMA. During the course of our study, we have observed that current methods for an effective understanding of differential splicing events within the transcriptomic landscape at high resolution are insufficient. To address this problem, we developed a computational model which has a potential to precisely estimate the “transcript expression levels” within a given gene locus by disentangling mature and nascent transcription contributions for each transcript at per base resolution. We modeled exonic and intronic read coverages by applying a non-linear computational model and estimated expression for each transcript, which best approximated the observed expression in total RNA-Seq data. The performance of our model was good in terms of computational processing time and memory usage. The application of our model is in the detection of differential splicing events. At exon level, differences in the ratio of the sum of mature and the sum of nascent transcripts over all the transcripts in a gene locus gives an indication of differential splicing. We have implemented our model in R-statistical language.