An analytic approach for interpretable predictive models in high-dimensional data in the presence of interactions with exposures

Blood transcriptome differentiates clinical clusters for asthma

Background

In previous studies, several asthma phenotypes were identified using clinical and demographic parameters. Transcriptional phenotypes were mainly identified using sputum and bronchial cells.

Objective

We aimed to investigate asthma phenotypes via clustering analysis using clinical variables and compare the transcription levels among clusters using gene expression profiling of the blood.

Methods

Clustering analysis was performed using 6 parameters: age of asthma onset, body mass index, pack-years of smoking, forced expiratory volume in 1 s (FEV1), FEV1/forced vital capacity, and blood eosinophil counts. Peripheral blood mononuclear cells (PBMCs) were isolated from whole blood samples and RNA was extracted from selected PBMCs. Transcriptional profiles were generated (Illumina NovaSeq 6000) and analyzed using the reference genome and gene annotation files (hg19.refGene.gft). Pathway enrichment analysis was conducted using GO, KEGG, and REACTOME databases.

Results

In total, 355 patients with asthma were included in the analysis, of whom 72 (20.3%) had severe asthma. Clustering of the 6 parameters revealed 4 distinct subtypes. Cluster 1 (n = 63) had lower predicted FEV1 % and higher pack-years of smoking and neutrophils in sputum. Cluster 2 (n = 43) had a higher proportion and number of eosinophils in sputum and blood, and severe airflow limitation. Cluster 3 (n = 110) consisted of younger subjects with atopic features. Cluster 4 (n = 139) included features of late-onset mild asthma. Differentially expressed genes between clusters 1 and 2 were related to inflammatory responses and cell activation. Th17 cell differentiation and interferon gamma-mediated signaling pathways were related to neutrophilic inflammation in asthma.

Conclusion

Four clinical clusters were differentiated based on clinical parameters and blood eosinophils in adult patients with asthma form the Cohort for Reality and Evolution of Adult Asthma in Korea (COREA) cohort. Gene expression profiling and molecular pathways are novel means of classifying asthma phenotypes.

Operation of a TCA cycle subnetwork in the mammalian nucleus

Nucleic acid and histone modifications critically depend on the tricarboxylic acid (TCA) cycle for substrates and cofactors. Although a few TCA cycle enzymes have been reported in the nucleus, the corresponding pathways are considered to operate in mitochondria. Here, we show that a part of the TCA cycle is operational also in the nucleus. Using 13C-tracer analysis, we identified activity of glutamine-to-fumarate, citrate-to-succinate, and glutamine-to-aspartate routes in the nuclei of HeLa cells. Proximity labeling mass spectrometry revealed a spatial vicinity of the involved enzymes with core nuclear proteins. We further show nuclear localization of aconitase 2 and 2-oxoglutarate dehydrogenase in mouse embryonic stem cells. Nuclear localization of the latter enzyme, which produces succinyl-CoA, changed from pluripotency to a differentiated state with accompanying changes in the nuclear protein succinylation. Together, our results demonstrate operation of an extended metabolic pathway in the nucleus, warranting a revision of the canonical view on metabolic compartmentalization.

Parallel and nonparallel genomic responses contribute to herbicide resistance in Ipomoea purpurea, a common agricultural weed

The repeated evolution of herbicide resistance has been cited as an example of genetic parallelism, wherein separate species or genetic lineages utilize the same genetic solution in response to selection. However, most studies that investigate the genetic basis of herbicide resistance examine the potential for changes in the protein targeted by the herbicide rather than considering genome-wide changes. We used a population genomics screen and targeted exome re-sequencing to uncover the potential genetic basis of glyphosate resistance in the common morning glory, Ipomoea purpurea, and to determine if genetic parallelism underlies the repeated evolution of resistance across replicate resistant populations. We found no evidence for changes in 5-enolpyruvylshikimate-3-phosphate synthase (EPSPS), glyphosate's target protein, that were associated with resistance, and instead identified five genomic regions that showed evidence of selection. Within these regions, genes involved in herbicide detoxification-cytochrome P450s, ABC transporters, and glycosyltransferases-are enriched and exhibit signs of selective sweeps. One region under selection shows parallel changes across all assayed resistant populations whereas other regions exhibit signs of divergence. Thus, while it appears that the physiological mechanism of resistance in this species is likely the same among resistant populations, we find patterns of both similar and divergent selection across separate resistant populations at particular loci.

An analytic approach for interpretable predictive models in high-dimensional data in the presence of interactions with exposures

Predicting a phenotype and understanding which variables improve that prediction are two very challenging and overlapping problems in the analysis of high-dimensional (HD) data such as those arising from genomic and brain imaging studies. It is often believed that the number of truly important predictors is small relative to the total number of variables, making computational approaches to variable selection and dimension reduction extremely important. To reduce dimensionality, commonly used two-step methods first cluster the data in some way, and build models using cluster summaries to predict the phenotype. It is known that important exposure variables can alter correlation patterns between clusters of HD variables, that is, alter network properties of the variables. However, it is not well understood whether such altered clustering is informative in prediction. Here, assuming there is a binary exposure with such network-altering effects, we explore whether the use of exposure-dependent clustering relationships in dimension reduction can improve predictive modeling in a two-step framework. Hence, we propose a modeling framework called ECLUST to test this hypothesis, and evaluate its performance through extensive simulations. With ECLUST, we found improved prediction and variable selection performance compared to methods that do not consider the environment in the clustering step, or to methods that use the original data as features. We further illustrate this modeling framework through the analysis of three data sets from very different fields, each with HD data, a binary exposure, and a phenotype of interest. Our method is available in the eclust CRAN package.