Bioinformatics Seminar
Time:
Venue: Na
30 May 2017
NaEvaluating RNA-seq differential expression at the transcript level
Gordon SmythWEHI Bioinformatics
In most of our routine RNA-seq analyses we have found it more profitable to analyse RNA-seq differential expression at the gene level rather than at the transcript level. The recent arrival of very fast light-weight transcript quantification software (kallisto and Salmon) is however likely to make transcript-level analyses of RNA-seq far more popular worldwide. Transcript level quantifications suffer very seriously though from read assignment uncertainty ; whereby reads that overlap multiple transcripts for the same gene cannot be unambiguously assigned. The level of uncertainty varies greatly between genes and transcripts and therefore needs to be predicted and built into the differential analysis procedures ; otherwise statistical analyses will be inefficient and possibly biased. This talk will describe how transcript-level differential expression analyses can be done efficiently with the limma and edgeR software packages using output from kallisto or Salmon.