Poster Presentation 23rd Annual Lorne Proteomics Symposium 2018

PepWitch: a python module to facilitate FDR-corrected Benjamini–Hochberg t-tests by employing same/same analysis on label-free shotgun data. (#100)

David C.L. Handler 1 , Flora Cheng 1 , Dana Pascovici 2 , Paul A Haynes 1
  1. Molecular Sciences, Macquarie University, Sydney, NSW, Australia
  2. Bioinformatics, APAF, Sydney, NSW, Australia

Label-free shotgun proteomics results often contain thousands of protein identifications. As such, employing the Student’s T-Test to ascertain differences between sample states will invariably introduce false positives. This is termed the multiple testing problem. In the recent past, multiple testing corrections such as the Bonferroni or Benjamini–Hochberg procedure were employed to account for this noise. However, these tests tend to be overly conservative while also eliminating true positive results. In our research, we have developed a new method to help correct the BenjaminiHochberg overcorrection through the analysis of the noise inherent within biological replicates. Through a multivariate, non-redundant analysis of six replicates (termed same/same analysis), an internal noise threshold is produced where an appropriate Q value facilitates a one percent false discovery rate. This Q value is then used for subsequent control/treated state comparisons. This research lead to the development of a python script called PepWitch (available soon on github) that can, given at least one state with six replicates, automatically determine the best Q value to use for subsequent control/treated analysis (GPM output only, for now). PepWitch also functions to generate high stringency data based on normalised spectral abundance factors from regular triplet experiments – this is an expansion of the previously available Scrappy module .


  1. Neilson, Karlie A., et al. "Label-free quantitative shotgun proteomics using normalized spectral abundance factors." Proteomics for Biomarker Discovery (2013): 205-222.