Winston Haynes, PhD CandidateDate:
2206 A&B Student Center
Understanding disease through integrated molecular and clinical analyses
Abstract:Traditional biomedical experiments are designed to study a single cohort for a single disease using a single technology. By studying disease with such a narrow lens, researchers make discoveries that are not reproducible because they are not representative of the real heterogeneity of disease. By integrating data from over 40 studies and 7,000 patients, we establish a robust signature of disease which correlates with disease activity and persists across blood, tissue, and sorted cell populations. We compare relationships of 104 diseases based on molecular and clinical manifestations from 41,000 gene expression samples and 2 million patient records. Finally, we contextualize single-cell RNA-seq data with bulk gene expression profiles to understand the relationships of novel cell subsets to known cell populations and human disease. By integrating biomedical datasets, my work has enabled an unbiased and multi-scale understanding of disease.
Bio: Winston Haynes is a PhD candidate in biomedical informatics at Stanford University. His research focuses on developing methods to improve understanding of disease through unbiased analyses of heterogeneous, publicly available data. Building off his discovery that publications are biased towards well-annotated genes instead of those with the strongest disease associations, his work integrates molecular and clinical evidence to identify overlooked aspects of disease, including therapeutically actionable relationships between seemingly disparate diseases and novel molecular pathways associated with disease activity
Epigenetic adaptation to environment in long lives trees
Oak represents a valuable natural resource across Northern Hemisphere ecosystems, attracting a large research community studying its genetics, ecology, conservation, and management. Here we introduce a draft genome assembly of valley oak (Quercus lobata) using Illumina, PacBio and Dovetail sequencing of adult leaf tissue of a tree found in an accessible, well-studied, natural southern California population. We next utilize this genome to carry out landscape epigenetics studies. DNA methylation in plants affects transposon silencing, transcriptional regulation and thus phenotypic variation. One unanswered question is whether DNA methylation could be involved in local adaptation of plant populations to their environments. If methylation alters phenotypes to improve plant response to the environment, then methylation sites or the genes that affect them could be a target of natural selection. Using reduced-representation bisulphite sequencing (RRBS) data, we assessed whether climate is associated with variation in DNA methylation levels among 58 naturally occurring, and species-wide samples of valley oak (Quercus lobata) collected across climate gradients. Environmental association analyses revealed 43 specific loci that are significantly associated with any of four climate variables, the majority of which are associated with mean maximum temperature. The 43 climate-associated SMVs tend to occur in or near genes, several of which have known involvement in plant response to environment. Multivariate analyses show that climate and spatial variables explain more overall variance epigenetic than genetic marks. Together, these results from natural oak populations provide initial evidence for a role of CG methylation in locally adaptive evolution or plasticity in plant response
2206A Student Center
Blockchain Technologies for Healthcare – Technical Challenges and Potential Applications
Blockchain is a technology of distributed ledger originally applied in the financial world. Bitcoin is one of the most adoptable cryptocurrencies based on Blockchain technique. The success of Bitcoin in technology means Blockchain has the potential for decentralized transaction validation, data provenance, data sharing, and data integration in different fields. Ethereum is a Blockchain-based platform with smart contract functionality inside. Smart contract is similar to coded protocol which enforce the workflow of data sharing. To date, most of academic papers for Blockchain in non-financial domains are still very conceptual and creating skepticism about the applicability of the technology and what it can achieve. At MU, we are experiencing the Ethereum platform to develop informatics tools for applications in health care, such as health information exchange, recruitment for clinical trials, etc. One of our scenarios is that through the Blockchain patients can share their medical data, including EHR, PACS, and mobile device data, with other authorized facilities or institutions, smart contract could access patients’ data from several data providers and send to approved parties without jumping through the hoops. Patients can also change the data sharing setting at their fingertips using smart contracts. However, both data security and integrity are challenging using the Blockchain technology. In this talk, I will introduce the Blockchain concepts, implementation of Blockchain nodes, design of smart contracts for data access simulations, and discuss several project ideas in artificial intelligence and precision medicine.
2206A Student Center
Data Mining for Genetic Combinations Relevant to Autism Subtypes
Autism is characterized by a complex set of behavioral, social, and cognitive deficits. Extensive variation of these phenotypes suggests the existence of autism subtypes that likely have distinct genetic etiologies. The lack of unifying genotypes common to autism patients supports this subtype structure, and suggests that the onset of autism is due to combinations of genetic factors. The ability to precisely diagnose autism subtypes using genetic markers would lead to earlier and more specific treatments and improve outcomes, stressing the need for research which increases our understanding of the genetic etiologies of autism subtypes. In this research, we identify combinations of genetic factors that are associated with groups of autism patients with unifying behavioral profiles, yielding candidate genes to be investigated for their role in the development of these potential autism subtypes. Utilizing methods that combine bioinformatics strategies with data mining practices, we pursue three goals: the discovery of genetic combinations associated to a disease subgroup, the exploration of disease subgroups to find potential subtypes, and the analysis of relationships between genes and subgroups to identify relevant functional interactions.
2206A Student Center
Analysis of Influence of Additional Diagnostic Clues During Pathology Diagnosis
Traditional pathology diagnostic process routinely relies on disease-specific diagnostic clues. We propose an informatics pipeline to identify and quantify additional diagnostic clues that, in addition to traditional disease-specific clues, can improve diagnostic outcomes and decrease the chance of diagnostic pitfalls. We used our PathEdEx whole-slide imaging platform to record user activities related to diagnosing a cancerous tissue slide along with the biological features that were noted in the tissue by the examining pathologist as relevant to the diagnosis. To identify and quantify additional diagnostic clues that can improve diagnosis, we extended association rule mining techniques to measure information gain of the additional diagnostic clues. To validate our findings, we computed Kullback-Leibler divergence that indicates information gain generated by additional diagnostic clues.
« Previous 1 2 3 4 5 … 15 Next »