airpot 0.8.3

AI rescoring for extreme-scale proteomics

airpot is a library for AI rescoring of PSMs in extreme-scale proteomics experiments.

Installation

This library requires Python 3.8+ and can be installed with pip:

pip install airpot

Basic Usage

Using airpot requires that you collect a dataset of PSMs, which can easily be accomplished with the wheely-mammoth library, which is installed when you install airpot:

>>> from wheely.mammoth.parsers import read_encyclopedia_features
>>> ds = read_encyclopedia_features("data/*.features.txt")

You can then train a model using the brew function:

>>> from airpot import brew
>>> res = brew(ds)
>>> res
<airpot.backends.mokapot.MokapotResult object at 0x7f2f5e9b7850>
>>> res.psms.data.limit(5).toPandas()[[res.psms.peptide_column, *res.psms.score_columns, res.psms.target_column, res.psms.qvalue_column]]
                   sequence  mokapot score  target  mokapot q-value
0    -.LSLEGDHSTPPSAYGSVK.-       2.851714    True         0.002123
1      -.IMDPNIVGSEHYDVAR.-       2.554996    True         0.002123
2     -.VAQPTITDNKDGTVTVR.-       2.232809    True         0.002123
3  -.TNVNGGAIALGHPLGGSGSR.-       2.189039    True         0.002123
4           -.ASIHEAWTDGK.-       2.145052    True         0.002123
>>> for m in res.models:
...   print(m)
... 
A trained mokapot.model.Model object:
	estimator: LinearSVC(class_weight={0: 10, 1: 1}, dual=False, random_state=7)
	scaler: StandardScaler()
	features: ['primary', 'xCorrLib', 'xCorrModel', 'LogDotProduct', 'logWeightedDotProduct', 'sumOfSquaredErrors', 'weightedSumOfSquaredErrors', 'numberOfMatchingPeaks', 'numberOfMatchingPeaksAboveThreshold', 'averageAbsFragmentDeltaMass', 'averageFragmentDeltaMasses', 'isotopeDotProduct', 'averageAbsParentDeltaMass', 'averageParentDeltaMass', 'eValue', 'deltaRT', 'numMissedCleavage', 'pepLength']
A trained mokapot.model.Model object:
	estimator: LinearSVC(class_weight={0: 10, 1: 10}, dual=False, random_state=7)
	scaler: StandardScaler()
	features: ['primary', 'xCorrLib', 'xCorrModel', 'LogDotProduct', 'logWeightedDotProduct', 'sumOfSquaredErrors', 'weightedSumOfSquaredErrors', 'numberOfMatchingPeaks', 'numberOfMatchingPeaksAboveThreshold', 'averageAbsFragmentDeltaMass', 'averageFragmentDeltaMasses', 'isotopeDotProduct', 'averageAbsParentDeltaMass', 'averageParentDeltaMass', 'eValue', 'deltaRT', 'numMissedCleavage', 'pepLength']
A trained mokapot.model.Model object:
	estimator: LinearSVC(class_weight={0: 0.1, 1: 0.1}, dual=False, random_state=7)
	scaler: StandardScaler()
	features: ['primary', 'xCorrLib', 'xCorrModel', 'LogDotProduct', 'logWeightedDotProduct', 'sumOfSquaredErrors', 'weightedSumOfSquaredErrors', 'numberOfMatchingPeaks', 'numberOfMatchingPeaksAboveThreshold', 'averageAbsFragmentDeltaMass', 'averageFragmentDeltaMasses', 'isotopeDotProduct', 'averageAbsParentDeltaMass', 'averageParentDeltaMass', 'eValue', 'deltaRT', 'numMissedCleavage', 'pepLength']