Computational Aspects of Biological Information 2017

Speaker: Nancy Cox

We have applied the PrediXcan method in which we use GTEx as a reference to impute transcript levels for genes and test the association of those imputed transcript levels with ~1800 phenome codes developed for phenome-wide association studies. I will illustrate how we are using the resulting gene x medical phenome catalog to extend discovery research into identification of primary disease mechanism as well as novel approaches to translation of discoveries.

Speaker: Polina Anikeeva

Mammalian nervous system contains billions of neurons that exchange electrical, chemical and mechanical signals. Our ability to study this complexity is limited by the lack of technologies available for interrogating neural circuits across their diverse signaling modalities without inducing a foreign-body reaction. My talk will describe neural interface strategies pursued in my group aimed at mimicking the materials properties and transduction mechanisms of the nervous system. Specifically, I will discuss (1) Fiber-based probes for multifunctional interfaces with the brain and spinal cord circuits; (2) Magnetic nanotransducers for minimally invasive neural stimulation; and (3) Active scaffolds for neural tissue engineering and interrogation.

Fiber-drawing methods can be applied to create multifunctional polymer-based probes capable of simultaneous electrical, optical, and chemical probing of neural tissues in freely moving subjects. Similar engineering principles enable ultra-flexible miniature fiber-probes with geometries inspired by nerves, which permit simultaneous optical excitation and recording of neural activity in the spinal cord allowing for optical control of lower limb movement. Furthermore, fiber-based fabrication can be extended to design of scaffolds that direct neural growth and activity facilitating repair of damaged nerves.

Molecular mechanisms of action potential firing inspire the development of materials-based strategies for direct manipulation of ion transport across neuronal membranes. For example, hysteretic heat dissipation by magnetic nanomaterials can be used to remotely trigger activity of neurons expressing heat-sensitive ion channels. Since the alternating magnetic fields in the low radiofrequency range interact minimally with the biological tissues, the magnetic nanoparticles injected into the brain can act as transducers of wireless magnetothermal deep brain stimulation. Similarly, local hysteretic heating allows magnetic nanoparticles to disrupt protein aggregates associated with neurodegenerative disorders.

Speaker: Yoseph Barash

In this talk, I will review some of the challenges and advances we have made in building a new view of transcriptome complexity, along with and the kind of tools that can be used to infer it.

Speaker: Bernat Olle

Medicines that modulate the human microbiome hold enormous promise for treatment of immune and infectious diseases, as highlighted by the success of randomized, controlled trials of fecal transplantation for treatment of C. difficile infections and IBD, as well as early pilot data in GvHD. They also pose some unique scientific, translational, manufacturing, and regulatory challenges. Drugs based on defined bacterial consortia (characterized communities of limited numbers of live organisms) could potentially retain some of the beneficial attributes of fecal transplants while being amenable to development as pharmaceutical-grade agents. I will present examples of Vedanta’s efforts to develop first-in-class drugs based on rationally selected bacterial consortia for applications in immune and infectious diseases.

Speaker: Gill Bejerano

I will highlight some methodologies and examples of finding causal mutations in both medical genomics and comparative genomics contexts.

Speaker: Daniel MacArthur

Rapid advances in sequencing technology have led to the generation of genome-scale DNA sequencing data for more than 2 million individuals worldwide. These data represent incredibly powerful information about the distribution and impact of genetic variation, but major challenges remain to aggregating and harmonizing them. In this presentation I will describe the development of the Exome Aggregation Consortium (ExAC) and Genome Aggregation Database (gnomAD) databases, which combined represent exome and genome sequencing data for over 135,000 individuals. I will discuss approaches to analyzing genome data at massive scale, and the applications of these data to understanding human variation and gene function.

Speaker: Nir Yoseph

Human immunity relies on coordinated responses of ensembles of cells from various types and functional states. Over the past several decades, substantial work has been done to catalog the cell types, states, and interactions that inform these behaviors. The emerging field of single cell genomics has the potential to make important contributions to this body of work, by providing a way to systematically map previously unknown subsets of cells, predict their functionality, and ultimately modulate their relative abundance so as to favorably tune the emerging immune response. In this talk I will present our recent work, which explored these ideas within the context of HIV-1 infection, using single-cell RNA-Seq of dendritic cells from human donors. Through a combination of computational analysis and experimental validation we were able to identify a highly functional subset of anti-viral dendritic cells that is capable of effectively priming T cell responses in vitro. Furthermore, by integrating knowledge from existing genomics databases we were able to identify and test immunomodulators that increase the relative abundance of this functional subset in-vitro, thus potentially leading to a more potent anti-viral response.

While data sets produced by single cell genomics hold unique information, they also pose unique challenges, from idiosyncratic technical limitations to difficulty of interpretability. During this this talk, I will also describe some of the computational tools developed in my group, aiming to address these challenges and realize the potential of single-cell genomics.

Speaker: Todd Golub

It is now becoming possible to bring powerful genomic methods to bear on the process of cancer drug discovery. This talk will illustrate 1) the use of the Connectivity Map, a large-scale effort to develop a compendium of gene expression signatures of cellular state, 2) the use of PRISM, a new multiplexed, high throughput approach to screening cancer cell lines for anti-cancer activity and resistance, and 3) the use of cancer models (cell lines and patient-derived xenografts) in discovery, with particular focus on clonal drift in these models and the impact of that drift on drug sensitivity.

Speaker: Susan Holmes

Longitudinal studies using perturbations enable us to study the resilience of the human microbiome. These are particularly informative in the case of antibiotic courses and colonic clean-out. The heterogeneity of the sources of information (time, phylogenetic trees, community networks, samples, batches, noise levels, sequences) pose real visualization and analytic challenges that we have overcome using interactive multivariate mappings. These enable simple tree-aware projections and account for uncertainties using Bayesian MCMC implementations. Longitudinal studies using perturbations enable us to study the resilience of the human microbiome. These are particularly informative in the case of antibiotic courses and colonic clean-out. The heterogeneity of the sources of information (time, phylogenetic trees, community networks, samples, batches, noise levels, sequences) pose real visualization and analytic challenges that we have overcome using interactive multivariate mappings. These enable simple tree-aware projections and account for uncertainties using Bayesian MCMC implementations.

Speaker: Jonathan Carlson

Emerging and re-emerging infectious diseases are a central public health concern, with implications for economies and even the stability of nations. The current system for detecting pathogens is entirely reactive, reliant on aggregating local reports by clinicians, and typically takes weeks of symptomatic human infections before detection. Project Premonition aims to detect pathogens before they cause outbreaks, using mosquitoes to sample pathogens in the environment. Turning a mosquito into a device requires new technologies to autonomously locate, robotically collect, and computationally analyze mosquitoes and their genetic payload.  To this end, we have developed and field tested robotic smart traps that instantaneously classify insects based on wing beat frequencies and other features, allowing real-time monitoring and selective capture of flying insects of interest. The rich data collected from these traps provides unprecedented insight into mosquito foraging behavior, allowing us to predict where and when we’ll find mosquitos, which we expect will allow us to more rapidly locate disease while simultaneously improving the efficacy of mosquito control. Once mosquitos are captured, deep sequencing has the potential to provide insights into the population genetics of the mosquitos, the hosts on which they’ve feed, and pathogens in both mosquitos and hosts. This is essentially a pan metagenomics problem, in which source genetic material may be derived from anywhere on the tree of life, thus motivating very high throughput taxonomic classification methods that leverage all available sequences in public databases. In this talk, I will provide an overview of Project Premonition, results from field trials on the traps, and our algorithmic approach to making a scalable metagenomics pipeline.