Self-supervised clustering of mass spectrometry imaging data using contrastive learning.

Hang Hu,Jyothsna Padmakumar Bindu,Julia Laskin

doi:10.1039/d1sc04077d

Hang Hu, Jyothsna Padmakumar Bindu + Show 1 more

Open Access

https://doi.org/10.1039/d1sc04077d

Copy DOI

Journal: Chemical Science	Publication Date: Jan 1, 2022
Citations: 11	License type: CC BY 3.0

Affiliation: Purdue University West Lafayette

Abstract

Mass spectrometry imaging (MSI) is widely used for the label-free molecular mapping of biological samples. The identification of co-localized molecules in MSI data is crucial to the understanding of biochemical pathways. One of key challenges in molecular colocalization is that complex MSI data are too large for manual annotation but too small for training deep neural networks. Herein, we introduce a self-supervised clustering approach based on contrastive learning, which shows an excellent performance in clustering of MSI data. We train a deep convolutional neural network (CNN) using MSI data from a single experiment without manual annotations to effectively learn high-level spatial features from ion images and classify them based on molecular colocalizations. We demonstrate that contrastive learning generates ion image representations that form well-resolved clusters. Subsequent self-labeling is used to fine-tune both the CNN encoder and linear classifier based on confidently classified ion images. This new approach enables autonomous and high-throughput identification of co-localized species in MSI data, which will dramatically expand the application of spatial lipidomics, metabolomics, and proteomics in biological research.

Highlights

Mass spectrometry imaging (MSI) is a powerful label-free molecular imaging technique for biological research, which enables simultaneous localization of multiple classes of biomolecules with high sensitivity and unprecedented molecular speci city.[1,2,3,4] By acquiring a full mass spectrum in each pixel of a virtual grid, MSI generates hundreds of molecular images in a single experiment
We train a deep convolutional neural network (CNN) using MSI data from a single experiment without manual annotations to effectively learn high-level spatial features from ion images and classify them based on molecular colocalizations
The approach is based on training a CNN to learn representations of molecular localizations and classify molecular images into groups based on high-level spatial features

Summary

Introduction

Mass spectrometry imaging (MSI) is a powerful label-free molecular imaging technique for biological research, which enables simultaneous localization of multiple classes of biomolecules with high sensitivity and unprecedented molecular speci city.[1,2,3,4] By acquiring a full mass spectrum in each pixel of a virtual grid, MSI generates hundreds of molecular images in a single experiment. We train a deep convolutional neural network (CNN) using MSI data from a single experiment without manual annotations to effectively learn high-level spatial features from ion images and classify them based on molecular colocalizations. We introduce a robust self-supervised clustering approach, which enables efficient colocalization of molecules in individual MSI datasets by retraining a CNN and learning representations of highlevel molecular distribution features without annotations.

Results

Conclusion