BCB Graduate Seminar Nathan Johnson "Determining Rewiring Functional Effects of Alternative Splicing Variants on Protein-Protein Interactions" SL 402 Jan 18 12pm

Thursday, January 18, 2018
12:00 pm

"Determining Rewiring Functional Effects of Alternative Splicing Variants on Protein-Protein Interactions"


Nathan T. Johnson, PhD. Candidate

Bioinformatics and Computational Biology

Jan 18 SL 402 12pm


Alternative splicing (A.S.) of mRNA precursors provides an important regulatory mechanism and is a crucial step to the expression of 90-95% of all genes in human.  A.S. rearranges the key components of a gene (exons, introns, and untranslated regions (UTRs)) in order to provide new arrangements of RNA (transcripts) that are expected to modify 70% of protein functions.  The prevalence on which transcripts are A.S. has been demonstrated to be based on age, tissue, cell type, and disease state. However, while a number of computational methods exist to predict the functional impact of genetic variation, no methods exist that determine the impact of A.S. Here we are presenting a method developed to determine whether A.S.  affects a protein-protein interaction (PPIs). PPIs play a key role in cell functioning and have been linked to a growing number of complex genetic disorders. Wet lab approaches have laboriously determined some of these effects, but the estimated 200,000 A.S. transcripts effects on protein interaction are far from being cataloged.  One of the main challenges when predicting whether two proteins interact is that it requires sequence and structure information. Sequenced based methods benefit from great coverage as most proteins have a known sequence, but will most likely suffer from lower accuracy.  However, a method that includes structural and sequence information will be of higher accuracy, but will suffer from the limited proteins with resolved structures.  To solve this fundamental problem we have developed a methodology, which given a known protein interaction, whether through A.S. this interaction is modified.  We have incorporated a recently developed method Learning Under Privileged Information (LUPI) that allows the use of structure information if available to build the model, but only requires a sequence of the alternative spliced isoform and a reference protein interaction related to the alternative spliced isoform.  Our novel method achieves >88% accuracy and is expected to be an asset to the research community.