Accelerating the drug discovery process with mathematical 1 Accelerating the drug discovery process

  • View
    0

  • Download
    0

Embed Size (px)

Text of Accelerating the drug discovery process with mathematical 1 Accelerating the drug discovery process

  • 1

    Accelerating the drug discovery process with mathematical

    modelling and MATLAB

    David Gavaghan University of Oxford

    david.gavaghan@cs.ox.ac.uk

    Presenter Presentation Notes Director of the EPSRC DTC

    Head of Computational Biology Group which has grown from around 7 people when formed in 2005 to over 70 today including 7 faculty. Previously we were part of the Numerical Analysis Group in Oxford.

  • Overview • The increasing need for mathematical training in the life

    sciences – Predictive and quantitative modelling of biological systems

    • What is a Doctoral Training Centre? – The Oxford DTC Programmes – Role of MATLAB within our programmes

    • Development of the Chaste software package – Links to industry, particularly Pharma

    • Making Chaste available through MATLAB

    2

  • Multi-scale, multi-physics

    • Biological systems are

    – Multiscale: involve interdependent processes occurring on multiple spatial and temporal scales

    – Multiphysics: involve multiple physical mechanisms (transport, signalling, reaction….)

    – Complex: self-healing, adaptive, plastic, self-organising, reactive

    Presenter Presentation Notes Two key points.

    1. Physical processes and scales are intertwined. Processes cannot be decoupled and scales cannot be separated using our usual applied mathematics toolbox (asymptotics, perurbation theory, homogenisation etc). This has huge implications for the computational complexity of the modelling task.

    2. Systems can change themselves and adapt to changes in their environment in ways that may not be obvious.

  • Courtesy of Peter Kohl (Harefields)

    Normal beating Fibrillation

    Model complexity is dependent on the scientific questions being asked…….

    Pras Pathmanathan (FDA)

    • Scientific questions: what is the optimal shock strength to reverse this process? What is the precise effect of a drug?

    • Aiming for predictive, quantitative understanding in biology • Requires a new approach to scientific training

    Presenter Presentation Notes If we look at the previous slide, if we simply want to model the action of a drug on an ion channel, or the integrative effect of drug block at a cellular level, then our models can be quite simple (in these cases typically non-linear odes which can be solved very efficiently in MATLAB)

    If, however, we want to simulate a complex process such as the initiation of a life threatening arrhythmia (as shown in middle panel) the we may need to include all components from ion channel dynamics right up the full complexity of the cardiac geometry in our model.

    On left is normal beating and opening and closing of the heart valves. A diseased heart may go into fibrillation which is often a precursor to heart failure. How are processes such as fibrillation initiated and in particular how can drugs influence this process? We can only gain an understanding of these processes through an iterative interplay between experiment, mathematical modelling and very computationally expensive computer simulation using HPC resources.

    Movie on the right was generated on a large cluster (>1000 processors) and solves on a realistic heart geometry in the left ventricle with over 4million tetrahedral elements in the Finite Element mesh.

    Scientific questions we might ask

  • • Quantitative and Systems Pharmacology in the Post- genomic Era: New Approaches to Discovering Drugs and Understanding Therapeutic Mechanisms

    • An NIH White Paper by the QSP Workshop Group – October, 2011.

    • Peter K. Sorger (co-chair), Sandra R.B. Allerheiligen (co- chair), Darrell R. Abernethy, et al

    Why is this important to the Pharmaceutical industry?

    5

    • Quantitative and Systems Pharmacology in the Post- genomic Era: New Approaches to Discovering Drugs and Understanding Therapeutic Mechanisms

    • An NIH White Paper by the QSP Workshop Group – October, 2011.

    • Peter K. Sorger (co-chair), Sandra R.B. Allerheiligen (co- chair), Darrell R. Abernethy, et al

    Definition: The goal of QSP is to understand, in a precise, predictive manner, how drugs modulate cellular networks in space and time and how they impact human pathophysiology. QSP aims to develop formal mathematical and computational models that incorporate data at several temporal and spatial scales; these models will focus on interactions among multiple elements (biomolecules, cells, tissues etc.) as a means to understand and predict therapeutic and toxic effects of drugs.

    Definition: The goal of QSP is to understand, in a precise, predictive manner, how drugs modulate cellular networks in space and time and how they impact human pathophysiology. QSP aims to develop formal mathematical and computational models that incorporate data at several temporal and spatial scales; these models will focus on interactions among multiple elements (biomolecules, cells, tissues etc.) as a means to understand and predict therapeutic and toxic effects of drugs.

    The report also makes the overarching recommendation:

    Because industry has an acute need for trainees with strong skills in quantitative reasoning, network biology, and animal and human pharmacology, industry should

    be engaged in education as well as research.

    Presenter Presentation Notes In 2010 only 17 new compounds were approved for phase III clinical trials by the FDA – something needs to change…..

  • EPSRC Life Sciences Interface Doctoral Training Centres

    6

  • 7

    What is a Doctoral Training Centre? • Introduced by EPSRC’s Life Sciences Interface Programme in 2002

    • Address need for scientists capable of quantitative and predictive research in biological and medical sciences

    • Typically fund centre costs plus 5 cohorts each of ten students

    • Min of 25% taught training, strong emphasis on “transferable skills”

    • Oxford LSI DTC was one of the first two (other being in Edinburgh)

    • Three years ago rolled out across EPSRC portfolio (£300M) 2002 and 2012

    cohorts

  • 8

    The Oxford DTC Programmes

    • Three current Programmes – Life Sciences Interface (2002) – Systems Biology (2007) – Systems Approaches to Biomedical Science (2009, Industrial)

    • Industry partners include GSK, AZ, Roche, Novartis, Pfizer

    • Total of ~300 students to date, ~120 completed PhDs

    • ~25% go into industry, 75% into academic research

    Strong focus on mathematical training (regardless of background) largely facilitated by MATLAB

    Presenter Presentation Notes Explain differences between programmes

  • Use of MATLAB within the Programmes

    • Gain understanding and insight in the mathematical courses (basic to advanced)

    • Data analysis – basic graphical tool through to advanced statistical, image

    processing and data visualisation

    • Computational modelling – Prototyping through to research software

    • Toolboxes routinely used – Bioinformatics, Image processing, PDE, Statistics, [Parallel]

    9

  • Examples of MATLAB in DTC Research

    10

    Chris Arthurs and David Kay Adaptive p-refinement

    Tom Doel and Vicente Grau Lobe segmentation in the lung

    Presenter Presentation Notes Can’t get movies to run yet – can always skip….

  • The majority of computational modelling research in the DTC is done in MATLAB but…

    • Some problems are of a scale and complexity that bespoke scientific software must be developed

    – anatomically detailed multiscale, multiphysics problems such as whole-organ modelling

    – individual-based models bridging hybrid discrete/continuous, stochastic/deterministic such as cancer modelling

    11

    Presenter Presentation Notes The image on the right shows the 4-million element finite element mesh that was used in the simulation I showed earlier. It has embedded fibre architecture and vasculature. The systems of equations solved over the mesh for the electrical potential typically have around 30-60 unknowns at each node.

  • CHASTE: a MATLAB-inspired software development project

    12

    Presenter Presentation Notes Matlab is great but it can’t do everything……in particular it isn’t (yet) for the types of really complex problems such as the multi-scale, multi-physics research problems typically arising in systems biology and physiology and of increasing interest to the pharmaceutical industry, so, about 7 years ago we started to develop Chaste. Out goal at the outset was to make programming in Chaste as easy as programming in matlab so that we would get strong takeup from the community…..but it actually started as a training course within the DTC…..and has almost all been developed by DTC students (now post-docs)

  • What is Chaste?

    “Cancer Heart and Soft Tissue

    Environment”

    http://www.cs.ox.ac.uk/chaste/

    • Started in 2005 as a 4-week DTC course in Software Engineering

    • Library of Open Source (BSD) code for large- scale biological problems

    • Aim: produce a robust, extensible, reliable, re- usable and well-documented code base

    • Functionality