Annotation

Scripts

bash

annotation

Start here.

This script runs the full pipeline, and has options to perform each section individually. Most of the sections are not neccessary unless you wish to build your own model rather than using the one provided. If that is the case, it is highly recommended that an HPC cluster is used, as the process takes time.

get_data

Downloads the data files:

• Pfam-A.hmm - containing the hidden markov models from Pfam A

• enzyme.dat - containing the EC numbers and which proteins are members of those groups

• uniprot_sprot.fasta - containing the sequences of the proteins in swissprot

python

parse_result.py

Uses the output of hmmsearch (profile against sequence database) to generate a sparse matrix of hit scores between proteins and pfam hmms

identify.py

Produces a sparse, boolean matrix with Trues where a protein is annotated as having a particular EC number

calculate_stats.py

Calculates:

1. how much of the swissprot database does not have EC numbers
2. how many (and which) Pfam HMMs are not seen in swissprot
3. how many (and which) swissprot proteins do not have a family in pfam

ec_frequency.py

Plots a frequency bar chart for number of proteins per EC number, with logarithmic bar widths

pfam_plots.py

Plots frequency bar charts for:

• Number of hits per HMM

• Ratio of enzyme to non-enzyme hits per HMM

Modules

paths.py

A variable containing the path to the data folder

frequency_bars.py

A function to draw consistent bar charts

test_data.py

Functions to produce a test dataset

prep.py

Functions to:

• reduce the data by removing proteins that do not hit any HMMs

• remove a portion of the non enzyme proteins, as they will be less important to the learning process