Machine Learning In Genome Wide Association Studies

This eBook is a collection of articles from a Frontiers Research Topic. Frontiers Research Topics are very popular trademarks of the Frontiers Journals Series: they are collections of at least ten articles, all centered on a particular subject. With their unique mix of varied contributions from Original Research to Review Articles, Frontiers Research Topics unify the most influential researchers, the latest key findings and historical advances in a hot research area! Find out more on how to host your own Frontiers Research Topic or contribute to one as an author by contacting the Frontiers Editorial Office: frontiersin.org/about/contact.

"Genome-Wide Association Studies (GWAS) are a popular tool in statistical genomics that are used to identify genetic variants associated with various dis- eases. However, their success has been limited, in part because they typically do not incorporate interactions between variants to model target traits. Since Deep neural networks have been successful across domains abundant with com- plex signals, like speech, language, and vision, they are also popular candidates for modelling interactions between genetic variants. However, their black-box nature is a hindrance to their application for GWAS. In this thesis, we present a pipeline to train and interpret feedforward neu- ral networks to conduct a genome-wide association study (GWAS). We show that trained deep neural networks can be interpreted using feature-importance techniques to accurately distinguish and rank simulated causal genetic variants. We improve its accuracy by extending the pipeline to the multi-task setting, wherein we simultaneously model two related, simulated traits. We demon- strate the accuracy, reliability, and scalability of our approach by identifying most known Diabetes genetic risk factors found using a conventional GWAS on the UK Biobank"--

The study of Single Nucleotide Polymorphisms (SNPs) associated with human diseases is important for identifying pathogenic genetic variants and illuminating the genetic architecture of complex diseases. A Genome-wide association study (GWAS) examines genetic variation in different individuals and detects disease related SNPs. The traditional machine learning methods always use SNPs data as a sequence to analyze and process and thus may overlook the complex interacting relationships among multiple genetic factors. In this thesis, we propose a new hybrid deep learning approach to identify susceptibility SNPs associated with colorectal cancer. A set of SNPs variants were first selected by a hybrid feature selection algorithm, and then organized as 3D images using a selection of space-filling curve models. A multi-layer deep Convolutional Neural Network was constructed and trained using those images. We found that images generated using the space-filling curve model that preserve the original SNP locations in the genome yield the best classification performance. We also report a set of high risk SNPs associate with colorectal cancer as the result of the deep neural network model.

Genome-wide association studies (GWAS) have led to a great number of new findings in human genetics and genetic epidemiology. GWAS identifies DNA sequence variations using human genome data and identifies the genetic risk factors for common diseases. There are many challenges that remain when mapping the complex underlying relationships between genotypes and phenotypes in GWAS. Here, we attempt to improve the power to detect correct mapping in GWAS for disease prevention and treatment. We examine a number of assumptions in GWAS that have been made over the past decade, which need to be updated and discussed in light of recent GWAS algorithm development. To achieve this goal, we discuss some of the current assumptions of GWAS and all possible factors that could affect predictive power. Using simulation studies, we show statistical evidence of how different factors, including sample size, heritability, model misspecification, and measurement error, affect the power to detect correct genetic associations. These data have the potential to improve the design of GWAS. As epistasis is the key to studying GWAS, we specifically studied epistasis, which is believed to account for part of the missing heritability. To detect interactions, we developed permuted Random Forest (pRF), a scale-free method, which is based on the traditional machine learning method Random Forest (RF). This method accurately detects single nucleotide polymorphism (SNP)-SNP interactions and top interacting SNP pairs by estimating how much the power of a random forest classification model is influenced by removing pairwise interactions. We systematically tested this approach on a simulation study with datasets possessing various genetic constraints including heritability, number of SNPs, and sample size. Our methodology shows high success rates for detecting interacting SNP pairs. We also applied our approach to two bladder cancer datasets, which shows results consistent with well-studied methodologies and we built permuted Random Forest networks (PRFN), in which we used nodes to represent SNPs and edges to indicate interactions. Data suggest the pRF method could improve detection of pure gene-gene interactions. Classic methods used to detect genetic association in GWAS involved separating biological knowledge from genetic information, thus wasting useful biological information when modeling associations between genotypes and phenotypes. We therefore further developed a biological information guided machine learning methodology, based on Encyclopedia of DNA Elements (ENCODE), called ENCODE information guided synthetic feature Random Forest (E-SFRF). Instead of studying biological associations at the SNP level, we separated SNPs based on ENCODE information and grouped them into a particular gene or enhancer to calculate the synthetic feature (SF) on a higher level. In our study, we focused on genes or enhancers from the AHR pathway, which is involved in cancer development. This work showed that the E-SFRF method could identify consistent main effect models based on SFs from two independent bladder cancer studies. We further studied the SNP-SNP interactions inside the top main effect SFs and discovered interesting SNP-SNP interactions that may lead to strong main effects. We believe our method could increase the possibility of replicating results across different GWAS datasets by increasing both the consistency and accuracy in genetic studies. Overall, we have found that studying interactions among SNPs is essential to increasing the power to uncover genetic architectures. By developing different machine learning methods, pRF, and further incorporating biological information to develop E-SFRF, we were able to detect pure gene-gene interactions in a scale-free and non-parametric way, helping to increase repeatability and reliability of GWAS using biological knowledge.

Artificial intelligence (AI) is taking on an increasingly important role in our society today. In the early days, machines fulfilled only manual activities. Nowadays, these machines extend their capabilities to cognitive tasks as well. And now AI is poised to make a huge contribution to medical and biological applications. From medical equipment to diagnosing and predicting disease to image and video processing, among others, AI has proven to be an area with great potential. The ability of AI to make informed decisions, learn and perceive the environment, and predict certain behavior, among its many other skills, makes this application of paramount importance in today's world. This book discusses and examines AI applications in medicine and biology as well as challenges and opportunities in this fascinating area.

This book was written by soybean experts to cluster in a single publication the most relevant and modern topics in soybean breeding. It is geared mainly to students and soybean breeders around the world. It is unique since it presents the challenges and opportunities faced by soybean breeders outside the temperate world.

This book constitutes the refereed proceedings of the 5th European Conference on Evolutionary Computation, Machine Learning and Data Mining in Bioinformatics, EvoBIO 2007, held in Valencia, Spain, April 2007. Coverage brings together experts in computer science with experts in bioinformatics and the biological sciences. It presents contributions on fundamental and theoretical issues along with papers dealing with different applications areas.

With the development of next-generation sequencing technologies, we can detect numerous genetic variants associated with many diseases or complex traits over the past decades. Genome-wide association studies (GWAS) have been one of the most effective methods to identify those variants. It discovers disease-associated variants by comparing the genetic information between controls and cases. This approach is simple and effective and has been used by many studies. Before performing GWAS, we need to detect the genetic variants of the sample population. A subset of these variants, however, may have poor sequencing quality due to limitations in NGS or variant callers. In genetic studies that analyze a large number of sequenced individuals, it is critical to detect and remove those variants with poor quality as they may cause spurious findings. Here, I will present ForestQC, an efficient statistical tool for performing quality control on variants identified from NGS data by combining a traditional filtering approach and a machine learning approach, which outperforms widely used methods by considerably improving the quality of variants to be included in the analysis. Once this association is identified, the next step is to understand the genetic mechanism of rare variants on how the variants influence diseases, especially whether or how they regulate gene expression as they may affect diseases through gene regulation. However, it is challenging to identify the regulatory effects of rare variants because it often requires large sample sizes and the existing statistical approaches are not optimized for it. To improve statistical power, I will introduce a new approach, LRT-q, based on a likelihood ratio test that combines effects of multiple rare variants in a nonlinear manner and has higher power than previous approaches. I apply LRT-q to the GTEx dataset and find many novel biological insights. Recent studies have shown that omics data can be used for automatic disease diagnosis with machine learning algorithms. I will introduce an accurate and automated machine learning pipeline for the diagnosis of atopic dermatitis (AD) based on transcriptome and microbiota data. I will demonstrate that this classifier can accurately differentiate subjects with AD and healthy individuals. It also identifies a set of genes and microorganisms that are predictive for AD. I will show that they are directly or indirectly associated with AD.