Methods of Profiling Mass Spectral Data Using Neural Networks

Abstract

Methods are provided to classify and identify features in mass spectral data using neural network algorithms. A convolutional neural network (CNN) was trained to identify amino acids from an unknown protein sample. The CNN was trained using known peptide sequences to predict amino acid presence, diversity, and frequency, peptide length, subsequences of amino acids classified by features include aliphatic/aromatic, hydrophobic/hydrophilic, positive/negative charge, and combinations thereof. Mass spectra data of a sample unknown to the trained CNN was discretized into a one-dimensional vector and input into the CNN. The CNN models can potentially be integrated to determine the complete peptide sequence from a spectrum, thereby improving the yield of identifiable protein sequences from mass spec analysis.

What is claimed is: 
     
         1 . A method of identifying features in mass spectral data, comprising:
 training a convolutional neural network by inputting a first mass spectrum matched to an amino acid sequence into the convolutional neural network to produce a trained convolutional neural network;   obtaining from a mass spectrometer a second mass spectrum of a protein sample having an unknown amino acid sequence;   discretizing the second mass spectrum into a weighted vector;   inputting the weighted vector into the trained convolutional neural network; and   determining, by an output of the trained convolutional neural network, a predicted amino acid sequence corresponding to the second mass spectrum.   
     
     
         2 . The method of  claim 1 , further comprising, prior to inputting the first mass spectrum into the convolutional neural network, discretizing the first mass spectrum into a first weighted vector, wherein the first weighted vector corresponds to a peak height in segments of the first mass spectrum. 
     
     
         3 . The method of  claim 1 , wherein the weighted vector corresponds to a peak height in segments of the second mass spectrum. 
     
     
         4 . A method of identifying features in mass spectral data, comprising:
 training a convolutional neural network by inputting a mass spectra from a known protein sample and a corresponding known amino acid sequence into the convolutional neural network to produce a trained convolutional neural network;   obtaining a mass spectra of an unknown protein sample;   inputting the mass spectra of the unknown protein sample into the trained convolutional neural network; and   determining, by a first output of the trained convolutional neural network, a presence or absence of an amino acid in the unknown protein sample.   
     
     
         5 . The method of  claim 4 , further comprising determining, by a second output of the trained convolutional neural network, a length of a peptide sequence of the unknown protein sample. 
     
     
         6 . The method of  claim 4 , further comprising determining, by a third output of the trained convolutional neural network, a frequency of the amino acid in the peptide sequence of the unknown protein sample. 
     
     
         7 . The method of  claim 4 , further comprising discretizing the mass spectra of the unknown protein sample into a one-dimensional vector prior to inputting the mass spectra of the unknown protein sample into the trained convolutional neural network. 
     
     
         8 . The method of  claim 7 , wherein the one-dimensional vector corresponds to a presence or absence of a peak in each segment of the mass spectra of the unknown protein sample. 
     
     
         9 . The method of  claim 4 , further comprising, prior to inputting the mass spectra from the known protein sample into the convolutional neural network, discretizing the mass spectra of the known protein sample into a one-dimensional vector, wherein the one-dimensional vector corresponds to a presence or absence of a peak in each segment of the mass spectra of the known protein sample. 
     
     
         10 . The method of  claim 4 , further comprising, prior to inputting the mass spectra from the known protein sample into the convolutional neural network, discretizing the mass spectra of the known protein sample into a weighted vector, wherein the weighted vector corresponds to a peak height in each segment of the mass spectra of the known protein sample.