The slightly longer version

I work as an Application Scientist at Dotmatics where I configure our applications for clients who range from pharmaceutical companies to materials science processing sites. Previous to this, I worked as a Data Analysis Scientist in Machine Vision, working as part of the Data Analysis group at Diamond Light Source, South Oxfordshire, UK. Prior to Diamond, I worked in a product design start-up that specialised in the Internet of Things. My work there was a mixture of writing RESTful APIs in Python, setting up an end-to-end testing framework with integrated hardware, consulting on the feasibility of new IOT projects and researching potential new clients. Before working in the startup, I completed an MSc. degree in Computer Science at the University of Bristol, which was a big change from my previous vocation as a research scientist in Chemical Biology.

My previous work at Diamond made use of deep learning, particularly convolutional neural networks, to create tools that help scientists who work at Diamond as well as those who visit as facility users. One of the main areas is segmentation of 3-dimensional image data such as that produced by X-ray computed tomography (XCT), electron cryo-tomography (cryoET) or serial block-face scanning electron microscopy (SBF-SEM). In addition, I have created deep learning tools that aid scientists who are performing experiments where they aim to grow protein crystals. For example, detecting the location of these crystals is the first step in collecting in-situ data on a fully automated macromolecular crystallography (MX) beamline at Diamond, known as VMXi.

Education

MSc in Computer Science

2016 | University of Bristol, UK

Thesis: Using Image Texture Analysis to Increase Protein Solubility and Detect Fluorescent Protein Crystals

Supervisors: Frank Von Delft and Julian Gough

PDF

About

After several years working in the laboratory on insecure, fixed-term, academic contracts I decided to take the plunge and change career. Having taught myself how to program in Python I left my job and enrolled in a Computer Science course in order to learn the fundamentals and gain more relevant experience.

Courses taken included: Programming in C and Java, Computer Architecture, Web Technologies and Databases. For my research project I set up a collaboration with a lab that I'd worked in previously. The project used texture analysis in combination with machine learning (clustering, random forests) to detect crystals in microscope images. Work was done in collaboration with the Structural Genomics Consortium, University of Oxford and iXpressGenes Ltd., Huntsville, Alabama.

Abstract

Growing a protein crystal is the first step in determining the 3D atomic structure of that protein using X-ray crystallography. Knowing the structure, in turn, gives insight into how these molecules function, how changes (mutations) such as those associated with cancers, affect their function and how drugs can be made to target specific proteins in order to treat disease. Growing a crystal is difficult and unlikely. To improve their chances, a crystallographer may set up hundreds of microscopic experiments. In each experiment, the protein of interest is mixed with a different ‘cocktail’ of chemical components that has been selected from a library of these cocktails. The experiments are then imaged using a robotic system and the micrographs either inspected manually or analysed computationally. TeXRank is software, developed by, and currently in use at, the Structural Genomics Consortium (SGC) at the University of Oxford. It uses texture analysis to score images for the likely presence of crystals and then displays these images to the experimenter in rank order. This reduces the number of images that have to be viewed before finding a crystal. This project builds upon this work with the aim of providing new tools to the experimenter that have the potential to improve their chances of successfully growing or identifying a crystal of a given protein. In particular, the project includes:

  • methodology to find a set of chemical cocktail conditions that can be used as a tool to improve the solubility of sparingly soluble proteins (a limiting factor in obtaining crystals)
  • a ‘virtual’ experiment to evaluate the potential usefulness of the solubility tool
  • adaptation of image processing and machine learning methods to allow TeXRank to use microscope images, captured under green light, of experiments containing protein labelled with a fluorescent dye
  • an evaluation of the usefulness of the novel combination of texture analysis with fluorescence labelling, both in comparison to standard imaging and to studies in the literature.
The main outcomes of this project were:
  • a set of 48 chemical cocktail conditions, selected through mining the results of image analysis on real experimental data and found to be significantly better as solubility tool than a random set of conditions (p < 0.05)
  • adaptations to the TeXRank processing pipeline and the training of a new classifier that allows the ranking of images from experiments containing fluorescent labelled protein
  • a comparison of ranking performance between pairs of images taken of the same experiments, proving that the use of fluorescence in combination with texture analysis is better than texture analysis alone
  • the fluorescence classifier created in this project also has superior ranking characteristics (area under receiver operating characteristic curve) to the current classifier used by TeXRank
  • comparison to the research literature determined that texton analysis was more accurate at detecting fluorescent protein crystals than a recent study that utilised image thresholding followed by edge detection

DPhil in Materials Science

2006 | University of Oxford, UK

Thesis: Studies on the Electrodeposition of Calcium Phosphate Coatings for Orthopaedic Applications and the Potential Incorporation of Apatite-Coated Liposomes

Supervisors: Jan T. Czernuszka

PDF

About

I was awarded funding by the Engineering and Physical Sciences Research Council for this independent research project to develop novel, improved coatings for orthopaedic implants with drug-delivery capability. Analysis techniques that I used included scanning and transmission electron microscopy, electron diffraction and X-ray powder diffraction crystallography.

When not in the lab, I spent time as the Chair of the departmental Joint Consultative Committee for Graduates and was the Social Secretary of St. Catherine's College Middle Common where I helped to organise parties and events for hundreds of graduate students. In addition I sat on the college IT committee and worked as a college porter and in a student IT support role.

Abstract

A low temperature electrodeposition technique has been used to coat metal substrates with calcium phosphate (CaP) phases at 37℃ in aqueous solution buffered at pH 7.4. Experiments have also been carried out on liposome vesicles, some populations of which were coated in CaP, the final aim being to combine the liposome vesicles with the coating methodology in order to produce composite coatings with a drug delivery capability.

Factors relating to the coating procedure that were altered included: the initial coating solution concentration, the drying method and the time for which the samples were allowed to remain in solution (maturation). The effects upon the coatings were studied using X-ray diffraction (XRD), low voltage-scanning electron microscopy (LV-SEM), transmission electron microscopy (TEM), electron diffraction and measurement of coating shear strengths.

LV-SEM revealed that one hour or more of maturation time in solution, after one electrodeposition coating cycle, led to a dramatic change in coating morphology from a thin particulate coating, to one with a coherent structure of interconnected crystallites. The initial coating cycle allowed the formation of a thin amorphous layer of CaP, which then acts as a seeding site for subsequent coating growth. TEM and electron diffraction of coatings revealed that the coating crystallinity increased with maturation time. The coating was determined to be hydroxyapatite (HAp) with a preferred orientation of crystallites with their c-axis perpendicular to the substrate that became more pronounced with increased maturation time.

Critical point drying (CPD) appeared to protect the structure of the coatings when compared to air drying. Shear strengths were modulated by alterations in maturation time. Penetration of the coating structure by epoxy resin during the testing procedure may have been a problem.

The results of these studies show that the treatment of the initial calcium phosphate layer deposited onto a metal substrate during coating has profound implications for coating structure and strength.

Lipid vesicles, manufactured from egg phosphatidylcholine (EPC) containing 30 mol % cholesterol and 4 mol % α-tocopherol, were characterised in terms of entrapment efficiency and release of both the marker molecule calcein and the antibiotic gentamicin. A higher percentage release of calcein was measured from CaP-coated liposomes when compared to uncoated vesicles when incubated for periods in excess of 70 h at 37℃, indicating that CaP ultimately increased the permeability of the liposomal membranes.

Addition of CaP-coated liposomes to coating solutions, during the electrodeposition of calcium phosphate, was found to moderate the decrease in HAp and octacalcium phosphate supersaturation that was normally seen in their absence. Composite coatings, incorporating liposomes that contained calcein were also manufactured.

BSc (Hons.) in Biology

2000 | University of Nottingham, UK

Thesis: The Effects of Autoregulated Cytokinin Production on the Physiology of Transgenic Tobacco and Lettuce Plants

Supervisors: Michael R. Davey

About

A broad-based Biology degree with courses taken in subjects such as Microbiology, Biochemistry, Bacterial Genes and Development, Plant Pathology, Plant Genetic Manipulation, Primatology and Animal Reproductive Physiology.

I specialised in Plant GM, with my dissertation written on "Using Transgenic Plants as Production Units for Proteinaceous Vaccines", and my final year lab-based project investigating how a gene for cytokinin production, regulated by a senescence- specific promoter, could help to increase the shelf life of lettuce plants in the supermarket. This was partly funded by a major supermarket chain, all before plant GM became unfashionable, of course.

I was also R&D Director for the virtual biotechnology company ‘ParaBond’, which progressed to the final of the Biotechnology YES (young entrepreneurs scheme) in the DTI, London. In addition I was responsible for organising speakers for the University Biology Society and I represented the university at cross-country running and small-bore rifle shooting.