Development of robust, single-subject markers of predictive inference for computational psychiatry

Lead supervisor: Professor Laurence Hunt

Co-supervisor: Professor Miriam Klein-Flügge

Commercial partner: P1vital

Several psychiatric and neurological disorders have been characterised in terms of alterations in decision making and predictive inference. There is widespread interest in understanding the neurophysiological basis of predictive inference – how the brain uses recent outcomes to make predictions about the probability of future events – which has been implicated in clinical conditions such as autism, obsessive-compulsive disorder, and schizophrenia.

However, a significant concern in studying clinically relevant individual differences within computational psychiatry has been the poor psychometric and neurometric properties of many commonly used tasks in the field. Identifying neural correlates that can be measured reliably in individual patients, within clinically feasible timescales, would be a key step toward biomarker development.

We have recently developed a continuous version of a widely used predictive inference task that is designed to enhance sensitivity to individual-level neural signals, enabling reliable estimation of belief-updating potentials from relatively brief neurophysiological (MEG/EEG) recordings. We have used test–retest analyses of neurophysiological data from participants performing this task in multiple sessions to demonstrate high reliability of individual ERP components associated with belief updating, using just 6-12 minutes of data per participant (Weber et al, in prep).

Building on this foundation, our proposed DPhil project has three aims:

(i) Analysis of recently collected data from a pharmacological study in patients with first episode psychosis. Patients have completed our task while randomised to either a standard treatment with a D2 receptor treatment for psychosis, or an adjunct treatment of D2 antipsychotic plus the NMDA receptor agonist memantine. The student will learn to analyse these data and apply these to this dataset, to study whether treatment with memantine affects neurophysiological signatures of predictive inference, in line with predictions from computational models of decision making (e.g. Cavanagh et al, eLife 2020);

(ii) Establish further neurophysiological indices of continuous tasks in other areas relevant to computational psychiatry (e.g. Facial Emotional Recognition Task (FERT) and/or Information Gathering Task), and establish test-retest reliability of these novel paradigms;

(iii) Collect comparative dataset on continuous tasks using the newly-installed, state-of-the-art optically pumped magnetometer (OPM)-MEG installation at the Warneford Hospital, to facilitate source-localisation of signals with high SNR, and evaluate OPM-MEG as a clinical tool.

A key element of the project will be the close involvement of our industry partners, P1vital, who will contribute throughout the project by advising on the clinical and commercial relevance of task design, digital biomarker validation, and patient-centred usability. In addition to this, the student will complete a 12-week placement at the company where they will lead an analysis project on data collected by P1vital, complementing their ongoing research projects in the rest of their DPhil.

Apply using course: DPhil in Experimental Psychology