A Bayesian framework that integrates multi-omics data and gene networks predicts risk genes from schizophrenia GWAS data

AbstractSchizophrenia (SCZ) is a polygenic disease with a heritability approaching 80%. Over 100 SCZ-related loci have so far been identified by genome-wide association studies (GWAS). However, the risk genes associated with these loci often remain unknown. We present a new risk gene predictor, rGAT-omics, that integrates multi-omics data under a Bayesian framework by combining the Hotelling and Box–Cox transformations. The Bayesian framework was constructed using gene ontology, tissue-specific protein–protein networks, and multi-omics data including differentially expressed genes in SCZ and controls, distance from genes to the index single-nucleotide polymorphisms (SNPs), and de novo mutations. The application of rGAT-omics to the 108 loci identified by a recent GWAS study of SCZ predicted 103 high-risk genes (HRGs) that explain a high proportion of SCZ heritability (Enrichment = 43.44 and $$p = 9.30 \times 10^{ - 9}$$ p = 9.30 × 1 0 − 9 ). HRGs were shown to be significantly ($$p_{\mathrm{adj}} = 5.35 \times 10^{ - 7}$$ p adj = 5.35 × 1 0 − 7 ) enriched in genes associated with neurological activities, and more likely to be expressed in brain tissues and SCZ-associated cell types than background genes. The predicted HRGs included 16 novel genes not present in any existing databases of SCZ-associated genes or previously predicted to be SCZ risk genes by any other method. More importantly, 13 of these 16 genes were not the nearest to the index SNP markers, and them would have been difficult to identify as risk genes by conventional approaches while ten out of the 16 genes are associated with neurological functions that make them prime candidates for pathological involvement in SCZ. Therefore, rGAT-omics has revealed novel insights into the molecular mechanisms underlying SCZ and could provide potential clues to future therapies.

Download Full-text

Comparison between instrumental variable and mediation-based approaches for reconstructing causal gene networks in yeast

Molecular Omics ◽

10.1039/d0mo00140f ◽

2021 ◽

Author(s):

Adriaan-Alexander Ludl ◽

Tom Michoel

Keyword(s):

Instrumental Variable ◽

Gene Networks ◽

Omics Data ◽

Causal Gene ◽

Causal Networks ◽

Transcriptomics Data ◽

A Cell ◽

Flow Of Information

Causal gene networks model the flow of information within a cell. Reconstructing causal networks from omics data is challenging because correlation does not imply causation. When genomics and transcriptomics data...

Download Full-text

“Guilt by association” is not competitive with genetic association for identifying autism risk genes

Scientific Reports ◽

10.1038/s41598-021-95321-y ◽

2021 ◽

Vol 11 (1) ◽

Author(s):

Margot Gunning ◽

Paul Pavlidis

Keyword(s):

Machine Learning ◽

Genetic Association ◽

Gene Networks ◽

Rare Variants ◽

Association Studies ◽

Genetic Disorders ◽

Autism Spectrum ◽

Biological Data ◽

Disease Genes ◽

Risk Genes

AbstractDiscovering genes involved in complex human genetic disorders is a major challenge. Many have suggested that machine learning (ML) algorithms using gene networks can be used to supplement traditional genetic association-based approaches to predict or prioritize disease genes. However, questions have been raised about the utility of ML methods for this type of task due to biases within the data, and poor real-world performance. Using autism spectrum disorder (ASD) as a test case, we sought to investigate the question: can machine learning aid in the discovery of disease genes? We collected 13 published ASD gene prioritization studies and evaluated their performance using known and novel high-confidence ASD genes. We also investigated their biases towards generic gene annotations, like number of association publications. We found that ML methods which do not incorporate genetics information have limited utility for prioritization of ASD risk genes. These studies perform at a comparable level to generic measures of likelihood for the involvement of genes in any condition, and do not out-perform genetic association studies. Future efforts to discover disease genes should be focused on developing and validating statistical models for genetic association, specifically for association between rare variants and disease, rather than developing complex machine learning methods using complex heterogeneous biological data with unknown reliability.

Download Full-text

Novel cancer subtyping method based on patient-specific gene regulatory network

10.1101/2021.03.24.436731 ◽

2021 ◽

Author(s):

Mai Adachi Nakazawa ◽

Yoshinori Tamada ◽

Yoshihisa Tanaka ◽

Marie Ikeguchi ◽

Kako Higashihara ◽

...

Keyword(s):

Gene Networks ◽

Regulatory Networks ◽

The Cancer Genome Atlas ◽

Patient Specific ◽

Specific Gene ◽

Omics Data ◽

Cancer Subtypes ◽

Molecular Systems ◽

Molecular Features ◽

Cancer Genome Atlas

The identification of cancer subtypes is important for the understanding of tumor heterogeneity. In recent years, numerous computational methods have been proposed for this problem based on the multi-omics data of patients. It is widely accepted that different cancer subtypes are induced by different molecular regulatory networks. However, only a few incorporate the differences between their molecular systems into the classification processes. In this study, we present a novel method to classify cancer subtypes based on patient-specific molecular systems. Our method quantifies patient-specific gene networks, which are estimated from their transcriptome data. By clustering their quantified networks, our method allows for cancer subtyping, taking into consideration the differences in the molecular systems of patients. Comprehensive analyses of The Cancer Genome Atlas (TCGA) datasets applied to our method confirmed that they were able to identify more clinically meaningful cancer subtypes than the existing subtypes and found that the identified subtypes comprised different molecular features. Our findings show that the proposed method, based on a simple classification using the patient-specific molecular systems, can identify cancer subtypes even with single omics data, which cannot otherwise be captured by existing methods using multi-omics data.

Download Full-text

Biological network inference from single-cell multi-omics data using heterogeneous graph transformer

10.1101/2021.10.31.466658 ◽

2021 ◽

Author(s):

Anjun Ma ◽

Xiaoying Wang ◽

Cankun Wang ◽

Jingxian Li ◽

Tong Xiao ◽

...

Keyword(s):

Deep Learning ◽

Single Cell ◽

Gene Networks ◽

Network Inference ◽

Lung Tumor ◽

Superior Performance ◽

Omics Data ◽

Learning Platform ◽

Cell Clustering ◽

Cell Cell

We present DeepMAPS, a deep learning platform for cell-type-specific biological gene network inference from single-cell multi-omics (scMulti-omics). DeepMAPS includes both cells and genes in a heterogeneous graph to infer cell-cell, cell-gene, and gene-gene relations simultaneously. The graph attention neural network considers a cell and a gene with both local and global information, making DeepMAPS more robust to data noises. We benchmarked DeepMAPS on 18 datasets for cell clustering and network inference, and the results showed that our method outperforms various existing tools. We further applied DeepMAPS on a case study of lung tumor leukocyte CITE-seq data and observed superior performance in cell clustering, and predicted biologically meaningful cell-cell communication pathways based on the inferred gene networks. To improve the feasibility and ensure the reproducibility of analyzing scMulti-omics data, we deployed a webserver with multi-functions and various visualizations. Overall, we valued DeepMAPS as a novel platform of the state-of-the-art deep learning model in the single-cell study and can promote the use of scMulti-omics data in the community.

Download Full-text

Identification of disease-associated loci using machine learning for genotype and network data integration

Bioinformatics ◽

10.1093/bioinformatics/btz310 ◽

2019 ◽

Vol 35 (24) ◽

pp. 5182-5190 ◽

Cited By ~ 4

Author(s):

Luis G Leal ◽

Alessia David ◽

Marjo-Riita Jarvelin ◽

Sylvain Sebert ◽

Minna Männikkö ◽

...

Keyword(s):

Machine Learning ◽

Gene Networks ◽

Association Studies ◽

R Package ◽

Biological Data ◽

Machine Learning Algorithms ◽

Supplementary Information ◽

Genome Wide Association Studies ◽

Omics Data ◽

Missing Heritability

Abstract Motivation Integration of different omics data could markedly help to identify biological signatures, understand the missing heritability of complex diseases and ultimately achieve personalized medicine. Standard regression models used in Genome-Wide Association Studies (GWAS) identify loci with a strong effect size, whereas GWAS meta-analyses are often needed to capture weak loci contributing to the missing heritability. Development of novel machine learning algorithms for merging genotype data with other omics data is highly needed as it could enhance the prioritization of weak loci. Results We developed cNMTF (corrected non-negative matrix tri-factorization), an integrative algorithm based on clustering techniques of biological data. This method assesses the inter-relatedness between genotypes, phenotypes, the damaging effect of the variants and gene networks in order to identify loci-trait associations. cNMTF was used to prioritize genes associated with lipid traits in two population cohorts. We replicated 129 genes reported in GWAS world-wide and provided evidence that supports 85% of our findings (226 out of 265 genes), including recent associations in literature (NLGN1), regulators of lipid metabolism (DAB1) and pleiotropic genes for lipid traits (CARM1). Moreover, cNMTF performed efficiently against strong population structures by accounting for the individuals’ ancestry. As the method is flexible in the incorporation of diverse omics data sources, it can be easily adapted to the user’s research needs. Availability and implementation An R package (cnmtf) is available at https://lgl15.github.io/cnmtf_web/index.html. Supplementary information Supplementary data are available at Bioinformatics online.

Download Full-text

Simultaneous inference of phenotype-associated genes and relevant tissues from GWAS data via Bayesian integration of multiple tissue-specific gene networks

Journal of Molecular Cell Biology ◽

10.1093/jmcb/mjx059 ◽

2017 ◽

Vol 9 (6) ◽

pp. 436-452 ◽

Cited By ~ 8

Author(s):

Mengmeng Wu ◽

Zhixiang Lin ◽

Shining Ma ◽

Ting Chen ◽

Rui Jiang ◽

...

Keyword(s):

Gene Networks ◽

Gwas Data ◽

Specific Gene ◽

Simultaneous Inference ◽

Tissue Specific ◽

Tissue Specific Gene ◽

Bayesian Integration

Download Full-text

Cell type-specific proteogenomic signal diffusion for integrating multi-omics data predicts novel schizophrenia risk genes

10.1101/2020.05.28.121517 ◽

2020 ◽

Author(s):

Abolfazl Doostparast Torshizi ◽

Jubao Duan ◽

Kai Wang

Keyword(s):

Single Cell ◽

Brain Regions ◽

Cellular Protein ◽

Cell Protein ◽

Tissue Cell ◽

Omics Data ◽

Cell Type ◽

Risk Genes ◽

The Brain ◽

Polygenic Diseases

AbstractAccumulation of diverse types of omics data on schizophrenia (SCZ) requires a systems approach to jointly modeling the interplay between genome, transcriptome and proteome. Proteome dynamics, as the definitive cellular machinery in human body, has been lagging behind the research on genome/transcriptome in the context of SCZ, both at tissue and single-cell resolution. We introduce a Markov Affinity-based Proteogenomic Signal Diffusion (MAPSD) method to model intra-cellular protein trafficking paradigms and tissue-wise single-cell protein abundances. MAPSD integrates multi-omics data to amplify the signals at SCZ risk loci with small effect sizes, and reveal convergent disease-associated gene modules in the brain interactome as well as more than 130 tissue/cell-type combinations. We predicted a set of high-confidence SCZ risk genes, the majority of which are not directly connected to SCZ susceptibility risk genes. We characterized the subcellular localization of proteins encoded by candidate SCZ risk genes in various brain regions, and illustrated that most are enriched in neuronal and Purkinje cells in cerebral cortex. We demonstrated how the identified gene set may be involved in different developmental stages of the brain, how they alter SCZ-related biological pathways, and how they can be effectively leveraged for drug repurposing. MAPSD can be applied to other polygenic diseases, yet our case study on SCZ signifies how tissue-adjusted protein-protein interaction networks can assist in generating deeper insights into the orchestration of polygenic diseases.

Download Full-text