Precision Psychiatry Through the Eye: Retinal Biomarkers for Predicting Lithium Response in Bipolar Disorder

Background

Bipolar disorder affects millions of people worldwide and is associated with substantial health inequalities, including reduced life expectancy, high rates of physical comorbidity, and barriers to timely access to effective treatment. Lithium remains the most effective mood stabiliser for many patients and is considered the gold-standard treatment for bipolar disorder. However, clinicians currently have no objective way to determine who will benefit from lithium before treatment begins. As a result, treatment selection often relies on trial-and-error approaches, delaying effective care and exposing some individuals to unnecessary side effects.

One of the major unmet challenges in precision psychiatry is the identification of biomarkers that can predict lithium response. While clinical response is routinely assessed using the validated Alda scale (Nunes et al., 2020), reliable biological predictors remain elusive. Progress has also been limited by an incomplete understanding of lithium’s mechanisms of action.

Recent advances in retinal neuroscience have opened an exciting new avenue for investigation. The discovery of the melanopsin-mediated light-sensitive retinal pathway (Berson et al., 2002), which contributes directly to circadian rhythm regulation, has led to the hypothesis that lithium may act by modulating abnormal retinal responses to light. This hypothesis is being investigated through the Wellcome Trust-funded Helios-BD study (Roguski et al., 2024). Retinal imaging and visual function workstreams within Helios-BD are led by the proposed supervisory team in Edinburgh, placing the student at the forefront of a highly novel and internationally competitive research programme.

Aims

Overall objective: To identify retinal biomarkers of lithium responsiveness in bipolar disorder and develop the foundations for objective, accessible tools that support personalised treatment selection in precision psychiatry.

The project will pursue four interrelated aims:

1. Determine whether retinal structure and function capture biological signatures of lithium responsiveness in bipolar disorder.

Using the Helios-BD dataset, assess associations between Alda scores and multimodal retinal measures, including optical coherence tomography (OCT), psychophysical assessments of cone-opponent pathway function, chromatic pupillometry, and visual evoked potentials (VEPs).

2. Define a multimodal retinal biomarker profile that distinguishes lithium responders from non-responders.

Identify individual and combined retinal biomarkers that are most strongly associated with treatment response, providing new insights into the neurobiological mechanisms underlying lithium efficacy.

3. Develop explainable predictive models for precision treatment stratification.

Apply advanced statistical and machine learning approaches to integrate multimodal retinal data and evaluate their potential to predict lithium responsiveness at the individual patient level, with a focus on clinically interpretable models.

4. Establish the foundations for retinal biomarkers as scalable tools for precision psychiatry.

Evaluate the potential for rapid, non-invasive retinal assessments to support equitable and personalised treatment selection in bipolar disorder, reducing reliance on trial-and-error prescribing and improving access to precision mental healthcare.

Training Outcomes

The student will receive interdisciplinary training spanning psychiatry, neuroscience, ophthalmic imaging, visual electrophysiology, and data science. Working with a uniquely rich multimodal dataset, they will develop expertise in retinal imaging biomarkers, psychophysical testing, signal processing, statistical modelling, and predictive analytics.

Training will include advanced analysis of OCT, pupillometry, electrophysiology, and psychophysical data; implementation of machine learning pipelines in Python; statistical analysis in R; and application of explainable AI methods for biomarker identification and validation. Through collaboration with experts in Informatics and clinical neuroscience, the student will gain experience in integrating complex biological datasets to address clinically important questions.

The project will equip the student with a highly sought-after combination of clinical, quantitative, and computational skills, positioning them to become a future leader in translational neuroscience, precision psychiatry, and biomarker-driven healthcare research.

References: 

Berson DM, Dunn FA, Takao M (2002). Phototransduction by retinal ganglion cells that set the circadian clock. Science. 295 (5557): 1070–1073.

Nunes A, Trappenberg T, Alda M (2020). international Consortium on Lithium Genetics (ConLiGen). Asymmetrical reliability of the Alda score favours a dichotomous representation of lithium responsiveness. PLoS One.15(1):e0225353.

Roguski A, Needham N, MacGillivray T, Martinovic J, Dhillon B, Riha RL, Armstrong L, Campbell IH, Ferguson A, Hilgen G, Lako M, Ritter P, Santhi N, von Schantz M, Spitschan M, Smith DJ (2024). Investigating light sensitivity in bipolar disorder (HELIOS-BD). Wellcome Open Res. 9:64.