Unique peptide-level statistics through spectral clustering – Amin Saffari – Shotgun proteomics background

View Github Repository Open presentation in a new window

khikho

See all presentation from khikho

Unique peptide-level statistics through spectral clustering – Amin Saffari – Shotgun proteomics background

0 0

SBC_UniqPeptide

Presentation about my project at Lukas group

On Github khikho / SBC_UniqPeptide

Unique peptide-level statistics through spectral clustering

Amin Saffari

Supervisor: Lukas Käll

Uppsala University

Agenda

  • Background
  • Motivation
  • Problem definition
  • Method
  • Results
  • Future works

Shotgun proteomics background

Target and Decoy search

  • Target DB: Theoratical spectra
  • Decoy DB: Simulation of incorrect spectra
  • PEP: Probability that a given peptide is incorrect

Motivation

  • Better score
  • Probability of unique peptides

Problem definition

Unique and not unique PSMs

  • Normally using the best score
  • No methods to combin these PSMs
  • These PSMs are not probabilistically independent to each other respect to their score (can't using Fisher's method)

Mehode

Clustering methode

  • Computing the area under the curve (trapozoid)
  • Binning the data

Clustering two spectra

Results

Score Before and After clustering

Target and Decoy PSMs score distribution

Conclusion & Future works

  • Succeeding on unique peptide level statistic (~90%)
  • Clustering without binning
  • Clustering based on the theoratical spectra of the respect peptide

Acknowledgements

Thank You!