SIP-MS leverages shotgun proteomics techniques to offer collagenous peptide-based species identification. It stands on two foundational pillars: a machine learning method classifier (a Random Forest classifier) with species-specific peptide sequences and abundances, and a correlation classifier that considers all informative peptides in a dataset.
SIP-MS is integrated as the back-end algorithm for the current SSE GUI, which, as of January 2024, encompasses 8 species. SSE takes into account various factors provided by SIP-MS results. It then either yields a prediction score for the submitted sample to identify the species or offers a similarity score for a sample in cases where SSE determines that the species for the submitted samples is not in the current database.
Here you can see the current species inventory
Currently, SSE can be utilized locally with the following command in R versions higher than 4.0:
install.packages("remotes")
install.packages("shiny")
library(shiny)
library(remotes)
install_github("hassanakthv/SIPMS")
library(SIPMS)
runGitHub(repo = "hassanakthv/SIPMS",subdir = "R")
The peptide file should have a column named "Peptide" as the first column and the rest of the columns should denote the samples names and contain the peptide abundances.
Here is the workflow for using the SSE interface: