Identification of integrated proteomics and transcriptomics signature of alcohol-associated liver disease using machine learning.

Distinguishing between alcohol-associated hepatitis (AH) and alcohol-associated cirrhosis (AC) remains a diagnostic challenge. In this study, we used machine learning with transcriptomics and proteomics data from liver tissue and peripheral mononuclear blood cells (PBMCs) to classify patients with alcohol-associated liver disease. The conditions in the study were AH, AC, and healthy controls. We processed 98 PBMC RNAseq samples, 55 PBMC proteomic samples, 48 liver RNAseq samples, and 53 liver proteomic samples. First, we built separate classification and feature selection pipelines for transcriptomics and proteomics data. The liver tissue models were validated in independent liver tissue datasets. Next, we built integrated gene and protein expression models that allowed us to identify combined gene-protein biomarker panels. For liver tissue, we attained 90% nested-cross validation accuracy in our dataset and 82% accuracy in the independent validation dataset using transcriptomic data. We attained 100% nested-cross validation accuracy in our dataset and 61% accuracy in the independent validation dataset using proteomic data. For PBMCs, we attained 83% and 89% accuracy with transcriptomic and proteomic data, respectively. The integration of the two data types resulted in improved classification accuracy for PBMCs, but not liver tissue. We also identified the following gene-protein matches within the gene-protein biomarker panels: CLEC4M-CLC4M, GSTA1-GSTA2 for liver tissue and SELENBP1-SBP1 for PBMCs. In this study, machine learning models had high classification accuracy for both transcriptomics and proteomics data, across liver tissue and PBMCs. The integration of transcriptomics and proteomics into a multi-omics model yielded improvement in classification accuracy for the PBMC data. The set of integrated gene-protein biomarkers for PBMCs show promise toward developing a liquid biopsy for alcohol-associated liver disease.

Integrated Transcriptomic and Proteomic Analysis Identifies Plasma Biomarkers of Hepatocellular Failure in Alcohol-Associated Hepatitis.

Alcohol-associated hepatitis (AH) is a form of liver failure with high short-term mortality. Recent studies have shown that defective function of hepatocyte nuclear factor 4 alpha (HNF4a) and systemic inflammation are major disease drivers of AH. Plasma biomarkers of hepatocyte function could be useful for diagnostic and prognostic purposes. Herein, an integrative analysis of hepatic RNA sequencing and liquid chromatography-tandem mass spectrometry was performed to identify plasma protein signatures for patients with mild and severe AH. Alcohol-related liver disease cirrhosis, nonalcoholic fatty liver disease, and healthy subjects were used as comparator groups. Levels of identified proteins primarily involved in hepatocellular function were decreased in patients with AH, which included hepatokines, clotting factors, complement cascade components, and hepatocyte growth activators. A protein signature of AH disease severity was identified, including thrombin, hepatocyte growth factor α, clusterin, human serum factor H-related protein, and kallistatin, which exhibited large abundance shifts between severe and nonsevere AH. The combination of thrombin and hepatocyte growth factor α discriminated between severe and nonsevere AH with high sensitivity and specificity. These findings were correlated with the liver expression of genes encoding secreted proteins in a similar cohort, finding a highly consistent plasma protein signature reflecting HNF4A and HNF1A functions. This unbiased proteomic-transcriptome analysis identified plasma protein signatures and pathways associated with disease severity, reflecting HNF4A/1A activity useful for diagnostic assessment in AH.

Differentiating between liver diseases by applying multiclass machine learning approaches to transcriptomics of liver tissue or blood-based samples.

Background & Aims: Liver disease carries significant healthcare burden and frequently requires a combination of blood tests, imaging, and invasive liver biopsy to diagnose. Distinguishing between inflammatory liver diseases, which may have similar clinical presentations, is particularly challenging. In this study, we implemented a machine learning pipeline for the identification of diagnostic gene expression biomarkers across several alcohol-associated and non-alcohol-associated liver diseases, using either liver tissue or blood-based samples. Methods: We collected peripheral blood mononuclear cells (PBMCs) and liver tissue samples from participants with alcohol-associated hepatitis (AH), alcohol-associated cirrhosis (AC), non-alcohol-associated fatty liver disease, chronic HCV infection, and healthy controls. We performed RNA sequencing (RNA-seq) on 137 PBMC samples and 67 liver tissue samples. Using gene expression data, we implemented a machine learning feature selection and classification pipeline to identify diagnostic biomarkers which distinguish between the liver disease groups. The liver tissue results were validated using a public independent RNA-seq dataset. The biomarkers were computationally validated for biological relevance using pathway analysis tools. Results: Utilizing liver tissue RNA-seq data, we distinguished between AH, AC, and healthy conditions with overall accuracies of 90% in our dataset, and 82% in the independent dataset, with 33 genes. Distinguishing 4 liver conditions and healthy controls yielded 91% overall accuracy in our liver tissue dataset with 39 genes, and 75% overall accuracy in our PBMC dataset with 75 genes. Conclusions: Our machine learning pipeline was effective at identifying a small set of diagnostic gene biomarkers and classifying several liver diseases using RNA-seq data from liver tissue and PBMCs. The methodologies implemented and genes identified in this study may facilitate future efforts toward a liquid biopsy diagnostic for liver diseases. Lay summary: Distinguishing between inflammatory liver diseases without multiple tests can be challenging due to their clinically similar characteristics. To lay the groundwork for the development of a non-invasive blood-based diagnostic across a range of liver diseases, we compared samples from participants with alcohol-associated hepatitis, alcohol-associated cirrhosis, chronic hepatitis C infection, and non-alcohol-associated fatty liver disease. We used a machine learning computational approach to demonstrate that gene expression data generated from either liver tissue or blood samples can be used to discover a small set of gene biomarkers for effective diagnosis of these liver diseases.

Association of Glyphosate Exposure with Blood DNA Methylation in a Cross-Sectional Study of Postmenopausal Women.

BACKGROUND: Glyphosate is the most commonly used herbicide in the world and is purported to have a variety of health effects, including endocrine disruption and an elevated risk of several types of cancer. Blood DNA methylation has been shown to be associated with many other environmental exposures, but to our knowledge, no studies to date have examined the association between blood DNA methylation and glyphosate exposure. OBJECTIVE: We conducted an epigenome-wide association study to identify DNA methylation loci associated with urinary glyphosate and its metabolite aminomethylphosphonic acid (AMPA) levels. Secondary goals were to determine the association of epigenetic age acceleration with glyphosate and AMPA and develop blood DNA methylation indices to predict urinary glyphosate and AMPA levels. METHODS: For 392 postmenopausal women, white blood cell DNA methylation was measured using the Illumina Infinium MethylationEPIC BeadChip array. Glyphosate and AMPA were measured in two urine samples per participant using liquid chromatography-tandem mass spectrometry. Methylation differences at the probe and regional level associated with glyphosate and AMPA levels were assessed using a resampling-based approach. Probes and regions that had an false discovery rate q<0.1 in ≥90% of 1,000 subsamples of the study population were considered differentially methylated. Differentially methylated sites from the probe-specific analysis were combined into a methylation index. Epigenetic age acceleration from three epigenetic clocks and an epigenetic measure of pace of aging were examined for associations with glyphosate and AMPA. RESULTS: We identified 24 CpG sites whose methylation level was associated with urinary glyphosate concentration and two associated with AMPA. Four regions, within the promoters of the MSH4, KCNA6, ABAT, and NDUFAF2/ERCC8 genes, were associated with glyphosate levels, along with an association between ESR1 promoter hypomethylation and AMPA. The methylation index accurately predicted glyphosate levels in an internal validation cohort. AMPA, but not glyphosate, was associated with greater epigenetic age acceleration. DISCUSSION: Glyphosate and AMPA exposure were associated with DNA methylation differences that could promote the development of cancer and other diseases. Further studies are warranted to replicate our results, determine the functional impact of glyphosate- and AMPA-associated differential DNA methylation, and further explore whether DNA methylation could serve as a biomarker of glyphosate exposure. https://doi.org/10.1289/EHP10174.

DEGnext: classification of differentially expressed genes from RNA-seq data using a convolutional neural network with transfer learning.

BACKGROUND: A limitation of traditional differential expression analysis on small datasets involves the possibility of false positives and false negatives due to sample variation. Considering the recent advances in deep learning (DL) based models, we wanted to expand the state-of-the-art in disease biomarker prediction from RNA-seq data using DL. However, application of DL to RNA-seq data is challenging due to absence of appropriate labels and smaller sample size as compared to number of genes. Deep learning coupled with transfer learning can improve prediction performance on novel data by incorporating patterns learned from other related data. With the emergence of new disease datasets, biomarker prediction would be facilitated by having a generalized model that can transfer the knowledge of trained feature maps to the new dataset. To the best of our knowledge, there is no Convolutional Neural Network (CNN)-based model coupled with transfer learning to predict the significant upregulating (UR) and downregulating (DR) genes from both trained and untrained datasets. RESULTS: We implemented a CNN model, DEGnext, to predict UR and DR genes from gene expression data obtained from The Cancer Genome Atlas database. DEGnext uses biologically validated data along with logarithmic fold change values to classify differentially expressed genes (DEGs) as UR and DR genes. We applied transfer learning to our model to leverage the knowledge of trained feature maps to untrained cancer datasets. DEGnext's results were competitive (ROC scores between 88 and 99[Formula: see text]) with those of five traditional machine learning methods: Decision Tree, K-Nearest Neighbors, Random Forest, Support Vector Machine, and XGBoost. DEGnext was robust and effective in terms of transferring learned feature maps to facilitate classification of unseen datasets. Additionally, we validated that the predicted DEGs from DEGnext were mapped to significant Gene Ontology terms and pathways related to cancer. CONCLUSIONS: DEGnext can classify DEGs into UR and DR genes from RNA-seq cancer datasets with high performance. This type of analysis, using biologically relevant fine-tuning data, may aid in the exploration of potential biomarkers and can be adapted for other disease datasets.

Association of mammographic density with blood DNA methylation.

BACKGROUND: Altered DNA methylation may be an intermediate phenotype between breast cancer risk factors and disease. Mammographic density is a strong risk factor for breast cancer. However, no studies to date have identified an epigenetic signature of mammographic density. We performed an epigenome-wide association study of mammographic density. METHODS: White blood cell DNA methylation was measured for 385 postmenopausal women using the Illumina Infinium MethylationEPIC BeadChip array. Differential methylation was assessed using genome-wide, probe-level, and regional analyses. We implemented a resampling-based approach to improve the stability of our findings. RESULTS: On average, women with elevated mammographic density exhibited DNA hypermethylation within CpG islands and gene promoters compared to women with lower mammographic density. We identified 250 CpG sites for which DNA methylation was significantly associated with mammographic density. The top sites were located within genes associated with cancer, including HDLBP, TGFB2, CCT4, and PAX8, and were more likely to be located in regulatory regions of the genome. We also identified differential DNA methylation in 37 regions, including within the promoters of PAX8 and PF4, a gene involved in the regulation of angiogenesis. Overall, our results paint a picture of epigenetic dysregulation associated with mammographic density. CONCLUSION: Mammographic density is associated with differential DNA methylation throughout the genome, including within genes associated with cancer. Our results suggest the potential involvement of several genes in the biological mechanisms behind differences in breast density between women. Further studies are warranted to explore these potential mechanisms and potential links to breast cancer risk.

Evidence for gene-smoking interactions for hearing loss and deafness in Japanese American families.

BACKGROUND: This study investigated the relationship between smoking and hearing loss and deafness (HLD) and whether the relationship is modified by genetic variation. Data for these analyses was from the subset of Japanese American families collected as part of the American Diabetes Association Genetics of Non-insulin Dependent Diabetes Mellitus study. Logistic regression with generalized estimating equations assessed the relationship between HLD and smoking. Nonparametric linkage analysis identified genetic regions harboring HLD susceptibility genes and ordered subset analysis was used to identify regions showing evidence for gene-smoking interactions. Genetic variants within these candidate regions were then each tested for interaction with smoking using logistic regression models. RESULTS: After adjusting for age, sex, diabetes status and smoking duration, for each pack of cigarettes smoked per day, risk of HLD increased 4.58 times (odds ratio (OR) = 4.58; 95% Confidence Interval (CI): (1.40,15.03)), and ever smokers were over 5 times more likely than nonsmokers to report HLD (OR = 5.22; 95% CI: (1.24, 22.03)). Suggestive evidence for linkage for HLD was observed in multiple genomic regions (Chromosomes 5p15, 8p23 and 17q21), and additional suggestive regions were identified when considering interactions with smoking status (Chromosomes 7p21, 11q23, 12q32, 15q26, and 20q13) and packs-per-day (Chromosome 8q21). CONCLUSIONS: To our knowledge this was the first report of possible gene-by-smoking interactions in HLD using family data. Additional work, including independent replication, is needed to understand the basis of these findings. HLD are important public health issues and understanding the contributions of genetic and environmental factors may inform public health messages and policies.

Influence of cobalamin scarcity on diatom molecular physiology and identification of a cobalamin acquisition protein.

Diatoms are responsible for ~40% of marine primary production and are key players in global carbon cycling. There is mounting evidence that diatom growth is influenced by cobalamin (vitamin B(12)) availability. This cobalt-containing micronutrient is only produced by some bacteria and archaea but is required by many diatoms and other eukaryotic phytoplankton. Despite its potential importance, little is known about mechanisms of cobalamin acquisition in diatoms or the impact of cobalamin scarcity on diatom molecular physiology. Proteomic profiling and RNA-sequencing transcriptomic analysis of the cultured diatoms Phaeodactylum tricornutum and Thalassiosira pseudonana revealed three distinct strategies used by diatoms to cope with low cobalamin: increased cobalamin acquisition machinery, decreased cobalamin demand, and management of reduced methionine synthase activity through changes in folate and S-adenosyl methionine metabolism. One previously uncharacterized protein, cobalamin acquisition protein 1 (CBA1), was up to 160-fold more abundant under low cobalamin availability in both diatoms. Autologous overexpression of CBA1 revealed association with the outside of the cell and likely endoplasmic reticulum localization. Cobalamin uptake rates were elevated in strains overexpressing CBA1, directly linking this protein to cobalamin acquisition. CBA1 is unlike characterized cobalamin acquisition proteins and is the only currently identified algal protein known to be implicated in cobalamin uptake. The abundance and widespread distribution of transcripts encoding CBA1 in environmental samples suggests that cobalamin is an important nutritional factor for phytoplankton. Future study of CBA1 and other molecular signatures of cobalamin scarcity identified here will yield insight into the evolution of cobalamin utilization and facilitate monitoring of cobalamin starvation in oceanic diatom communities.