Predictive models of cell state change in health and disease

Additional Supervisor: Prof Luigi Del Debbio (University of Edinburgh)

This project would suit an ambitious, fearless and scientifically inquisitive individual from a physics, mathematics, statistics, computer science (or similar) background. The project is designed to fuse the biomedical expertise of one supervisor (CPP) with the mathematical physics expertise of the other (LDD). The intersection between these different disciplines provides a fertile ground for new concepts and discoveries, and for gaining skills that are badly-needed in academia and industry.

Background

Currently, we describe a cell’s type or state from data acquired at single time points. Consequently, our biological understanding of cellular phenomena over time is limited, particularly once a cell is challenged or stimulated. Such predictions are essential for designing experiments that test the physiological validity of molecular and cellular mechanisms of disease cause and consequence. Accurate models of transcriptomic dynamics upon perturbation would foreshadow a conceptual and experimental revolution in cellular and molecular biology, and in cellular genetics.

Many algorithms have been developed to infer gene regulatory networks (GRNs), based on local Bayesian networks, information theory, clustering, state-space and regression. GRNs have also been inferred from time-course data and sparse data from single cells. Network edges represent correlations in the levels of RNA transcripts across samples or cells. Molecular relationships can be separated into those that are due to (i) transcriptional events, in which genes are activated or inhibited in common by a particular transcription factor, or (ii) post-transcriptional events, in which the abundances of multiple RNA transcripts are regulated concomitantly by microRNA(s) because of their competition for binding to these RNA transcripts. Surprisingly, we do not yet know the relative contributions of these two phenomena to normal physiology or to disease.

Aims

Years 1 & 2. The student will start by undertaking the requisite personalised training and by reviewing the available literature on GRN inference and the molecular and cellular explanations of these networks. Then together we will build Ising models with pairwise couplings between transcripts (Nguyen et al., Adv Physics, 2017) from available bulk RNA-Seq data sets (e.g. GTEx) and single cell data sets (e.g. Human Cell Atlas). The simplest form of this model involves discrete variables that can be in one of two states between only ‘adjacent’ transcripts. We will use Restricted Boltzmann Machines to solve the inverse Ising problem and find the pairwise couplings from data in this simple case, which we use as a sandbox for applying the latest learning techniques (see e.g. Fischer & Igel, 2012). We will then develop more complex models using continuous variables that represent couplings between all transcript pairs.

These models should effectively capture the data structure but will not, at this stage, provide robust predictions of regulatory interactions. To achieve this, we will take advantage of time-course data and predictions of microRNA-mRNA and transcription factor-DNA binding interactions, regressing out network interactions that reflect cell type fluctuations between samples. Furthermore, we will overlay predictions of causal trans-eQTL interactions (i.e. gene A variants alter mRNA B level) making use of normal human DNA variation. Finally, we will integrate the network with GWAS summary statistics to identify subnetworks contributing to trait variation and disease susceptibility (Mancuso et al., 2017).

Years 3 & 4. By perturbing the resulting networks we will then predict phenomena that will be tested experimentally. The precise experimental perturbations that will be applied need to be defined once the networks’ predictive potential and the then available experimental procedures become clear. Nevertheless, they are likely to involve: for transcriptional events, high-throughput CRISPR/Cas9 genome editing coupled to single-cell sequencing (e.g. Dixit et al., 2016), and/or for post-transcriptional events, antagomir screening to inhibit microRNAs. Such experiments would likely take advantage of experimental systems already set-up in the CPP lab. The ultimate aims are to model, in human cells, variants that causally alter disease susceptibility via (post-) transcriptional mechanisms.

This MRC programme is joint between the Universities of Edinburgh and Glasgow. You will be registered at the host institution of the primary supervisor detailed in your project selection.

All applications should be made via the University of Edinburgh, irrespective of project location:

http://www.ed.ac.uk/studying/postgraduate/degrees/index.php?r=site/view&id=919

Please note, you must apply to one of the projects and you are encouraged to contact the primary supervisor prior to making your application. Additional information on the application process if available from the link above.

For more information about Precision Medicine visit:

http://www.ed.ac.uk/usher/precision-medicine