Identification of cardiomyopathy-related core genes through human metabolic networks and expression data

Background Cardiomyopathy is a complex type of myocardial disease, and its incidence has increased significantly in recent years. Dilated cardiomyopathy (DCM) and ischemic cardiomyopathy (ICM) are two common and indistinguishable types of cardiomyopathy. Results Here, a systematic multi-omics integration approach was proposed to identify cardiomyopathy-related core genes that could distinguish normal, DCM and ICM samples using cardiomyopathy expression profile data based on a human metabolic network. First, according to the differentially expressed genes between different states (DCM/ICM and normal, or DCM and ICM) of samples, three sets of initial modules were obtained from the human metabolic network. Two permutation tests were used to evaluate the significance of the Pearson correlation coefficient difference score of the initial modules, and three candidate modules were screened out. Then, a cardiomyopathy risk module that was significantly related to DCM and ICM was determined according to the significance of the module score based on Markov random field. Finally, based on the shortest path between cardiomyopathy known genes, 13 core genes related to cardiomyopathy were identified. These core genes were enriched in pathways and functions significantly related to cardiomyopathy and could distinguish between samples of different states. Conclusion The identified core genes might serve as potential biomarkers of cardiomyopathy. This research will contribute to identifying potential biomarkers of cardiomyopathy and to distinguishing different types of cardiomyopathy.

Microbiome Research and Multi-Omics Integration for Personalized Medicine in Asthma

Asthma is a multifactorial inflammatory disorder of the respiratory system characterized by high diversity in clinical manifestations, underlying pathological mechanisms and response to treatment. It is generally established that human microbiota plays an essential role in shaping a healthy immune response, while its perturbation can cause chronic inflammation related to a wide range of diseases, including asthma. Systems biology approaches encompassing microbiome analysis can offer valuable platforms towards a global understanding of asthma complexity and improving patients’ classification, status monitoring and therapeutic choices. In the present review, we summarize recent studies exploring the contribution of microbiota dysbiosis to asthma pathogenesis and heterogeneity in the context of asthma phenotypes–endotypes and administered medication. We subsequently focus on emerging efforts to gain deeper insights into microbiota–host interactions driving asthma complexity by integrating microbiome and host multi-omics data. One of the most prominent achievements of these research efforts is the association of refractory neutrophilic asthma with certain microbial signatures, including predominant pathogenic bacterial taxa (such as Proteobacteria phyla, Gammaproteobacteria class, especially species from Haemophilus and Moraxella genera). Overall, despite existing challenges, large-scale multi-omics endeavors may provide promising biomarkers and therapeutic targets for future development of novel microbe-based personalized strategies for diagnosis, prevention and/or treatment of uncontrollable asthma.

Unsupervised learning of cross-modal mappings in multi-omics data for survival stratification of gastric cancer

Aims: This study presents a survival stratification model based on multi-omics integration using bidirectional deep neural networks (BiDNNs) in gastric cancer. Methods: Based on the survival-related representation features yielded by BiDNNs through integrating transcriptomics and epigenomics data, K-means clustering analysis was performed to cluster tumor samples into different survival subgroups. The BiDNNs-based model was validated using tenfold cross-validation and in two independent confirmation cohorts. Results: Using the BiDNNs-based survival stratification model, patients were grouped into two survival subgroups with log-rank p-value = 9.05E-05. The subgroups classification was robustly validated in tenfold cross-validation (C-index = 0.65 ± 0.02) and in two confirmation cohorts (E-GEOD-26253, C-index = 0.609; E-GEOD-62254, C-index = 0.706). Conclusion: We propose and validate a robust and stable BiDNN-based survival stratification model in gastric cancer.

Deep learning applications in single-cell omics data analysis

Deep learning (DL) is a branch of machine learning (ML) capable of extracting high-level features from raw inputs in multiple stages. Compared to traditional ML, DL models have provided significant improvements across a range of domains and applications. Single-cell (SC) omics are often high-dimensional, sparse, and complex, making DL techniques ideal for analyzing and processing such data. We examine DL applications in a variety of single-cell omics (genomics, transcriptomics, proteomics, metabolomics and multi-omics integration) and address whether DL techniques will prove to be advantageous or if the SC omics domain poses unique challenges. Through a systematic literature review, we have found that DL has not yet revolutionized or addressed the most pressing challenges of the SC omics field. However, using DL models for single-cell omics has shown promising results (in many cases outperforming the previous state-of-the-art models) but lacking the needed biological interpretability in many cases. Although such developments have generally been gradual, recent advances reveal that DL methods can offer valuable resources in fast-tracking and advancing research in SC.

The molecular pathophysiology of mood disorders: from the analysis of single molecular layers to multi-omic integration.

Next-generation sequencing now enables the rapid and affordable production of reliable biological data at multiple molecular levels, collectively referred to as “omics”. To maximize the potential for discovery, computational biologists have created and adapted integrative multiomic analytical methods. When applied to diseases with traceable pathophysiology such as cancer, these new algorithms and statistical approaches have enabled the discovery of clinically relevant molecular mechanisms and biomarkers. In contrast, these methods have been much less applied to the field of molecular psychiatry, although diagnostic and prognosticbiomarkers are similarly needed. In the present review, we first briefly summarize main findings from two decades of studies that investigated single molecular processes in relation to mood disorders. Then, we conduct a systematic review of multi-omic strategies that have been proposed and used more recently. We also list databases and types of data available to researchers for future work. Finally, we present the newest methodologies that have been employed for multi-omics integration in other medical fields, and discuss their potential for molecular psychiatry studies.

Supervised Hierarchical Autoencoders for Multi-Omics Integration in Cancer Survival Models

With the increasing amount of high-throughput sequencing data becoming available, the proper integration of differently sized and heterogeneous molecular and clinical groups of variables has become crucial in cancer survival models. Due to the difficulty of multi-omics integration, the Cox Proportional-Hazards (Cox PH) model using clinical data has remained one of the best-performing methods [Herrmann et al., 2021]. This motivates the need for new models which can successfully perform multi-omics integration in survival models and outperform the Cox PH model. Furthermore, there is a strong need to make multi-omics models more sparse and interpretable to encourage their usage in clinical settings. We developed a neural architecture, termed Supervised Hierarchical Autoencoder (SHAE), based on supervised autoencoders and Sparse-Group-Lasso regularization. Our new method performed competitively with the best performing statistical models used for multi-omics survival analysis. Moreover, it outperformed the Cox PH model using clinical data across all 17 cancers from The Cancer Genome Atlas (TCGA) considered in our work. We further showed that surrogate linear models for SHAE trained on a subset of multi-omics groups achieved competitive performance at consistently high sparsity levels, enabling usage within clinics. Alternatively, surrogate models can act as a feature selection step, permitting improved performance in arbitrary downstream survival models. Code for the reproduction of our results is available on Github.

eQTLs are key players in the integration of genomic and transcriptomic data for phenotype prediction

Multi-omics represent a promising link between phenotypes and genome variation. Few studies yet address their integration to understand genetic architecture and improve predictability. Our study used 241 poplar genotypes, phenotyped in two common gardens, with their xylem and cambium RNA sequenced at one site, yielding large phenotypic, genomic and transcriptomic datasets. For each trait, prediction models were built with genotypic or transcriptomic data and compared to concatenation integrating both omics. The advantage of integration varied across traits and, to understand such differences, we made an eQTL analysis to characterize the interplay between the genome and the transcriptome and classify the predicting features into CIS or TRANS relationships. A strong and significant negative correlation was found between the change in predictability and the change in predictor importance for eQTLs (both TRANS and CIS effects) and CIS regulated transcripts, and mostly for traits showing beneficial integration and evaluated in the site of transcriptomic sampling. Consequently, beneficial integration happens when redundancy of predictors is decreased, leaving the stage to other less prominent but complementary predictors. An additional GO enrichment analysis appeared to corroborate such statistical output. To our knowledge, this is a novel finding delineating a promising way to explore data integration.One-sentence summarySuccessful multi-omics integration when predicting phenotypes makes redundant the predictors that are linked to ubiquitous connections between the omics, according to biological and statistical approaches