Gradients of Generative Models for Improved Discriminative Analysis of Tandem Mass Spectra

TL;DR

This paper introduces a novel approach that uses gradients of generative models, specifically a dynamic Bayesian network, to enhance the discriminative analysis of tandem mass spectrometry data, leading to improved protein identification accuracy.

Contribution

It proposes a Fisher kernel based on the gradient of a DBN for better spectrum classification and introduces Theseus, a DBN that models multiple MS/MS scoring functions, enabling unsupervised parameter learning.

Findings

Fisher kernel significantly outperforms previous methods on tested datasets.

Theseus effectively models various MS/MS scoring functions.

Unsupervised learning of model parameters is achieved with gradient ascent and max-product inference.

Abstract

Tandem mass spectrometry (MS/MS) is a high-throughput technology used toidentify the proteins in a complex biological sample, such as a drop of blood. A collection of spectra is generated at the output of the process, each spectrum of which is representative of a peptide (protein subsequence) present in the original complex sample. In this work, we leverage the log-likelihood gradients of generative models to improve the identification of such spectra. In particular, we show that the gradient of a recently proposed dynamic Bayesian network (DBN) may be naturally employed by a kernel-based discriminative classifier. The resulting Fisher kernel substantially improves upon recent attempts to combine generative and discriminative models for post-processing analysis, outperforming all other methods on the evaluated datasets. We extend the improved accuracy offered by the Fisher kernel framework to other search algorithms by introducing Theseus, a DBN representing a large number of widely used MS/MS scoring functions. Furthermore, with gradient ascent and max-product inference at hand, we use Theseus to learn model parameters without any supervision.