Mark Dunning Bioinformatics Core Director, University of Sheffield

Statistical and Data Analysis Challenges in Bioinformatics

Bioinformatics is a multi-disciplinary subject that combines aspects of biology, computer science and statistics. Modern experimental techniques are able to generate vast amounts of data that can profile an individual's genome and offer insights into the development of disease and potential novel therapeutics. In this talk, I will describe the challenges faced by Bioinformaticians trying to deal with such data on a daily basis and the opportunities for collaboration with other disciplines to develop new analytical methods.

Tim Freeman PhD Student, University of Sheffield

Identifying genomic loci susceptible to systematic sequencing bias in clinical whole genomes

Accurate next-generation sequencing (NGS) of genetic variants is key to many areas of science and medicine, such as cataloguing population genetic variation and diagnosing genetic diseases. Certain genomic positions can be prone to higher rates of systematic sequencing and alignment bias that limit accuracy, resulting in false positive variant calls. Current standard practices to differentiate between loci that can and cannot be sequenced with high confidence utilise consensus between different sequencing methods as a proxy for sequencing confidence. These practices have significant limitations and alternative methods are required to overcome these.

We have developed a novel statistical method based on summarising sequenced reads from whole genome clinical samples and cataloguing them in “Incremental Databases” (IncDBs) that maintain individual confidentiality. Allele statistics were catalogued for each genomic position that consistently showed systematic biases with the corresponding NGS sequencing pipeline. We found systematic biases present at ~1-3% of the human autosomal genome across five patient cohorts. We identified which genomic regions were more or less prone to systematic biases, including large homopolymer flanks (OR=23.29-33.69) and the NIST high confidence genomic regions (OR=0.154-0.191). We confirmed our predictions on a gold-standard reference genome and showed that these systematic biases can lead to suspect variant calls within clinical panels.

Our results recommend increased caution to address systematic biases in whole genome sequencing and alignment. This study provides the implementation of a simple statistical approach to enhance quality control of clinically sequenced samples by flagging variants at suspect loci for further analysis or exclusion

Sorkatis Kariotis PhD Student, University of Sheffield

Cluster analysis of whole-blood gene expression to uncover heterogeneity of Idiopathic Pulmonary Arterial Hypertension

Idiopathic pulmonary arterial hypertension (IPAH) is a difficult to diagnose rare disease that has mostly unknown causes and describes a heterogeneous group of conditions. Its diagnosis can often be delayed by up to 3 years from the first symptom as the disease is defined by a diagnosis of exclusion of other forms of PAH which leads to a heterogeneous population of patients and a difficulty to define IPAH structure. To decipher the features that characterise a health condition or a clinical phenotype, this study selects two robust and consistent spectral clustering models utilizing RNA profiles to discriminate between finer independent sample subgroups. The clustering models extracted five pure IPAH and 4 mixed (IPAH/healthy) groups and the presence of significantly enriched clinical or genetic features was explored