Single-Cell Classification Using Mass Spectrometry through Interpretable Machine Learning

Abstract

The brain consists of organized ensembles of cells that exhibit distinct morphologies, cellular connectivity, and dynamic biochemistries that control the executive functions of an organism. However, the relationships between chemical heterogeneity, cell function, and phenotype are not always understood. Recent advancements in matrix-assisted laser desorption/ionization mass spectrometry have enabled the high-throughput, multiplexed chemical analysis of single cells, capable of resolving hundreds of molecules in each mass spectrum. We developed a machine learning workflow to classify single cells according to their mass spectra based on cell groups of interest (GOI), e.g., neurons vs astrocytes. Three data sets from various cell groups were acquired on three different mass spectrometer platforms representing thousands of individual cell spectra that were collected and used to validate the single cell classification workflow. The trained models achieved >80% classification accuracy and were subjected to the recently developed instance-based model interpretation framework, SHapley Additive exPlanations (SHAP), which locally assigns feature importance for each single-cell spectrum. SHAP values were used for both local and global interpretations of our data sets, preserving the chemical heterogeneity uncovered by the single-cell analysis while offering the ability to perform supervised analysis. The top contributing mass features to each of the GOI were ranked and selected using mean absolute SHAP values, highlighting the features that are specific to the defined GOI. Our approach provides insight into discriminating the chemical profiles of the single cells through interpretable machine learning, facilitating downstream analysis and validation.

Image-guided MALDI mass spectrometry for high-throughput single-organelle characterization

“…The most important features for the classification between LVs and DCVs (Fig. 3d ) were selected via Shapley additive explanations (SHAP) through a previously described method 6 , where a total of 97 features with nonzero mean SHAP values from the model output were selected, with 36 features putatively identified by mass-match assignment (Fig. 3e ).…”

“…Mass spectrometry (MS) imaging has begun being used for both cellular and some subcellular analyses in discovery-based studies 1 – 3 but is limited in throughput and spatial resolution for organelle measurements. Alternatively, both cells 4 – 6 and organelles 7 of interest can be isolated and placed on a glass slide for subsequent MS measurement. Though we used this approach to assay the peptides within individual organelles 7 , it was limited in throughput and difficult to automate due to the manual positioning of each individual organelle before measurement.…”

Interpretable Machine Learning for Genomics

“…Another SHAP variant -TreeExplainer (Lundberg et al, 2020), which is optimized for tree-based ensembles such as random forests -was used to identify taxa in the skin microbiome most closely associated with various phenotypic traits (Carrieri et al, 2021). SHAP is also gaining popularity in mass spectrometry, where data heterogeneity can complicate more classical inference procedures (Tideman et al, 2020;Xie et al, 2020). show that high white blood cell counts increase the negative risk conferred by high blood urea nitrogen for progression to end stage renal disease (ESRD).…”

Chemical Analysis of Single Cells and Organelles

“…Another publication by Sweedler and co-workers additionally elaborated on this data set which, along with others, was used in the development of a novel machine learning workflow for the classification of single cells into groups of interest (e.g., neurons vs astrocytes) according to their mass spectra. 130 …”