Dataset Information

A machine learning framework to predict cancer metabolomics from gene-expression data

ABSTRACT: Metabolomics provides a direct functional readout of a tumor’s physiology. Yet, it is lagging behind other omics technologies in facilitating disease monitoring and prognostication. This stems partly from the scarcity of large-scale metabolomic studies, but also the analytical complexities of detecting diverse metabolites with varying physicochemical properties and concentrations. To address this, we developed a machine learning framework using both tumor tissue and cell line samples across multiple cancer types that allows prediction of metabolomics from gene expression data. Two different model types were selected and trained for tissues and cell lines with their generalization capacity validated on independent cohorts, accurately predicting as high as 70-80% of tested metabolites. This work offers a scalable and efficient machine learning pipeline to determine metabolic from transcriptomic signatures, opening avenues to reconstruct and study the metabolic landscape of samples across novel and existing datasets lacking direct metabolomics measurements.

ORGANISM(S): Homo sapiens

PROVIDER: GSE299429 | GEO | 2025/09/30

REPOSITORIES: GEO

ACCESS DATA

Dataset's files

Source:

			Action	DRS
		Other

Items per page:

1 - 1 of 1

Similar Datasets

Project description:The interplay between pathogens and hosts has been studied for decades using targeted approaches such as the analysis of mutants and host immunological responses. Although much has been learned from such studies, they focus on individual pathways and fail to reveal the global effects of infection on the host. To alleviate this issue, high-throughput methods such as transcriptomics and proteomics have been used to study host-pathogen interactions. Recently, metabolomics was established as a new method to study changes in the biochemical composition of host tissues. We report a metabolomics study of Salmonella enterica serovar Typhimurium infection. We used Fourier Transform Ion Cyclotron Resonance Mass Spectrometry with Direct Infusion to reveal that dozens of host metabolic pathways are affected by Salmonella in a murine infection model. In particular, multiple host hormone pathways are disrupted. Our results identify unappreciated effects of infection on host metabolism and shed light on mechanisms used by Salmonella to cause disease, and by the host to counter infection. Female C57BL/6 mice were infected with Salmonella enterica serovar Typhimurium SL1344 cells by oral gavage. Feces and livers were collected and metabolites extracted using acetonitrile. For experiments with feces, samples were collected from 4 mice before and after infection. For liver experiments, 11 uninfected and 11 infected mice were used. Samples were combined into 3 groups of 3-4 mice each, resulting in the analysis of 3 group samples of uninfected and 3 of infected mice. Extracts were infused into a 12-T Apex-Qe hybrid quadrupole-FT-ICR mass spectrometer equipped with an Apollo II electrospray ionization source, a quadrupole mass filter and a hexapole collision cell. Raw mass spectrometry data were processed as described elsewhere (Han et al. 2008. Metabolomics. 4:128-140 [PMID 19081807]). To identify differences in metabolite composition between uninfected and infected samples, we filtered the list of masses for metabolites which were present on one set of samples but not the other. Additionally, we calculated the ratios between averaged intensities of metabolites from uninfected and infected mice. To assign possible metabolite identities, monoisotopic neutral masses of interest were queried against MassTrix (http://masstrix.org). Masses were searched against the Mus musculus database within a mass error of 3 ppm. Data were analyzed by unpaired t tests with 95% confidence intervals.

Project description:To cause disease, Salmonella enterica serovar Typhimurium requires two type-III secretion systems, encoded on Salmonella Pathogenicity Islands 1 and 2 (SPI-1 and -2). These secretion systems serve to deliver virulence proteins, termed effectors, into the host cell cytosol. While the importance of these effector proteins to promote colonization and replication within the host has been established, the specific roles of individual secreted effectors in the disease process are not well understood. In this study, we used an in vivo gallbladder epithelial cell infection model to study the function of the SPI-2-encoded effector, SseL. Deletion of the sseL gene resulted in bacterial filamentation and elongation and unusual localization of Salmonella within infected epithelial cells. Infection with the ?sseL strain also caused dramatic changes in lipid metabolism and led to massive accumulation of lipid droplets in infected cells. Some of these changes were investigated through metabolomics of gallbladder tissue. This phenotype was directly attributed to the deubiquitinase activity of SseL, as a Salmonella strain carrying a single point mutation in the catalytic cysteine resulted in the same phenotype as the deletion mutant. Excessive buildup of lipids due to the absence of a functional sseL gene was also observed in S. Typhimurium-infected livers. These results demonstrate that SseL alters host lipid metabolism in infected epithelial cells by modifying ubiquitination patterns of cellular targets. Female C57BL/6 mice were infected with the indicated strain of Salmonella enterica serovar Typhimurium by oral gavage. Four gallbladders were collected and pooled per sample group and metabolites extracted using a mixture of methanol and chloroform. Extracts were infused into a 12-T Apex-Qe hybrid quadrupole-FT-ICR mass spectrometer equipped with an Apollo II electrospray ionization source, a quadrupole mass filter and a hexapole collision cell. Raw mass spectrometry data were processed as described elsewhere (Han et al. 2008. Metabolomics. 4:128-140). To identify differences in metabolite composition between different groups of samples, we filtered the list of masses for metabolites which were present on one set of samples but not the other. Additionally, we calculated the ratios between averaged intensities of metabolites from each group of mice. To assign possible metabolite identities, monoisotopic neutral masses of interest were queried against MassTrix (http://masstrix.org). Masses were searched against the Mus musculus database within a mass error of 3 ppm.

Dataset Information

A machine learning framework to predict cancer metabolomics from gene-expression data

Dataset's files

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets