Deep learning enabling next-generation bacterial infection diagnostics from MALDI-ToF mass spectrometry

The project:
In time-critical infections such as severe sepsis, each hour without working antibiotics leads to a 6% increase in mortality. However, it often takes 48 hours to confirm clinical bacteraemia and a further 24 hours to determine antibiotic susceptibility. In the last decade, the use of Matrix Assisted Laser Desorption/Ionization Time-of-Flight (MALDI-ToF) mass spectrometry has cut analysis time down to a few seconds after a few hours/overnight culturing, and hence has made substantial clinical impact. Nevertheless, MALDI-ToF is reliant on informatics to first extract bacterial signals from the noisy spectral data, and then to reliably match these signals against a previously curated reference library of pathogens. Leading commercial software for MALDI-ToF, such as Bruker BioTyper, are able to identify >200 species. Nevertheless, they cannot differentiate all clinically-relevant species, are generally unable to determine sub-species, cannot determine antimicrobial susceptibility, and still require significant culturing time to achieve necessary accuracy. There is an urgent for new computational techniques to solve these key challenges and accelerate time-to-diagnosis.

Rather than considering just a small set of extracted peaks, our computational platform for mass spectrometry combines detailed modelling of the raw data and Bayesian statistics to perform robust difference discovery (http://www.biospi.org/). In this studentship, we will apply this platform to clinical MALDI-ToF for the first time. Moreover, based on statistical deep learning techniques, new methodology will be devised to model and separate bacterial and contaminant signals so that reference libraries can be automatically generated from large amounts of historical clinical data. From this we will develop a prediction model to probabilistically classify new acquisitions, validating performance on the breadth of clinical data and biobanked samples available to us.

This 3 year studentship must start in April 2018 at the latest. While funded by Veterinary Sciences and in collaboration with veterinarians, the studentship will be based within Prof Dowsey’s Data Science group in Bristol and will benefit from immersion and support from a multi-disciplinary team of both research and clinical microbiologists (Drs Matthew Avison and Maha Albur) and microbial geneticists (Profs Ed Feil and Sam Sheppard).

How to apply:
Please make an online application for this project at http://www.bris.ac.uk/pg-howtoapply. Please select ‘Faculty of Health Sciences’ and then ‘Veterinary Science_(PhD)’ on the Programme Choice page and enter details of the studentship when prompted in the Funding and Research Details sections of the form

Candidate requirements: The studentship would suit an applicant with a strong first degree or masters in a computational discipline (e.g. mathematics, computing, electrical engineering) and competent programming skills.

Contacts: Prof Andrew Dowsey ([Email Address Removed])