The Enhanced Prediction of Single-cell Metabolism through Consensus Methods
Date
2023-08-18
Authors
Amiss, John Michael
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
Motivation: Altered metabolism is a key contributor to pathology in multiple dis ease states. In cancers, metabolic shifts are crucial to providing the tumor cells
with the energy and substrates necessary for uncontrolled proliferation. Altered
metabolism can exist within certain pathological cells, or within a population of cells.
Unfortunately, direct measurements of single-cell metabolism are extremely limited.
However, single-cell flux prediction software can be used to infer cellular metabolism
from readily available transcriptome data. Two recent single-cell flux prediction tools,
Single-cell Flux Estimation Analysis (“scFEA”) and Compass, have been used in re search and have been validated by direct metabolite measurements. The two systems
make use of highly disparate methods. By adapting the outputs of these tools, I
hypothesized that they can work in concert to provide a higher confidence flux pre diction at multiple levels of resolution within a metabolic pathway.
Results: I developed a method to compare the outputs of Compass and scFEA and
to determine the patterns of agreement between the systems. By creating and testing
multiple methods of mapping metadata to reaction identifiers, and applying func tion composites to Compass outputs, I identified a method that showed the highest
general similarities with the outputs of scFEA across reaction clusters. I demon strated the application of this method to the results from running the two systems
on bulk and single-cell RNA-seq data. This analysis showed broad biological areas of
agreement between the results, including processes related to cellular energy produc tion, and both nucleotide and amino acid metabolism. Bulk and imputed data was
shown to sharply reduce the consensus between the systems. By creating a separate
reaction-level method, I validated my consensus approach across individual reactions.
This additionally revealed that the previous patterns of disagreement between the
systems spanned entire groups of related reactions. I tested the consensus measures
on two matched transcriptome and metabolomics datasets. The results suggest that
consensus between the two systems may indicate that both are coherent with direct
metabolic measurements.
Description
Keywords
Single-cell, Metabolism, Flux