Transcript Quantification with Kallisto

Skip the mapping step with Kallisto  

*Thanks to Anna Battenhouse to the text and figures!

Kallisto is a tool from the Pachter lab that performs quanitfication of transcripts without requiring alignment. 

  • No explicit alignment to reference genome or transciptome 
  • Instead, uses “pseudoalignment” to transcriptome 
    • For each read, determine not where in each transcript it aligns, but rather which transcripts it is compatible with 
    • Simultaneously addresses 2 aspects of “multi-mapping” reads in traditional RNAseq pipelines 
      • Multiple possible genomic loci (addressed during alignment) 
      • Multiple possible transcripts of origin (addressed during quantification) 
  • Pseudoalignments are sufficient to quantify transcript abundances 
    • Expectation Maximization (EM) algorithm is applied to a “simple” RNAseq Likelihood function 
    • Report estimated abundances as Transcripts per Million (TPM) + counts 

No P-value reported or differential expression (DE) support, but… 

  • kallisto re-runs EM on multiple bootstrap re-samples to estimate variance 
  • then kallisto bootstraps are used by add-on sleuth DE package

 

Why Kallisto?

Speed and performance are greatly improved with Kallisto.