Precision functional mapping for target engagement in drug development

Biomarkers are critical for advancing the development of experimental CNS therapeutics. Efforts to identify reliable drug biomarkers using resting and task fMRI have been underwhelming. Published fMRI studies have shown high variability, poor reproducibility, and lack of validated targets. This is a consequence of ignoring two critical principles 1) individual differences—individuals’ brains are structurally and functionally unique, and 2) measurement variability—reliable measurement of low-frequency fluctuations using fMRI requires 20–40 min high-quality data [1].

Precision functional mapping (PFM) overcomes many limitations of conventional resting fMRI through the use of dense repeated sampling, imaging advances (e.g., multi-band, multi-echo, Nordic denoising), and individual-specific network mapping and analysis [2]. PFM has revealed properties of brain organization and activity [1, 3] and phenotypes of neuropsychiatric illness [4] that are obscured in group-averaged data. On one side of the coin, personalized network mapping may reveal underpinnings of behavioral differences [5]. On the other, personalized network mapping makes it possible to control for individual differences, increasing the effect size of interventional and cohort studies. This application of PFM methodology has important implications to drug development.

Precision imaging drug trials (PIDTs) apply the advances of PFM to the study of psychoactive molecules. The first application of a precision imaging drug trial was somewhat serendipitous. When Laumann and colleagues began mapping functional networks in the brain of Russ Poldrack using the My Connectome dataset (N = 1), they noticed a large and reliable difference in brain connectivity between days when he was caffeinated or not [1]. The result was buried in a supplemental figure. But the conceptual advance was crucial—using a repeated sampling paradigm and personalized network mapping, we can measure large and reliable drug effects. Newbold and colleagues [3] beautifully adapted this approach to show that dominant arm casting disconnects inter-hemispheric motor circuits—reporting dramatic effect sizes in a small sample (N = 3).

We recently applied the PIDT approach to study the effects of psilocybin versus methylphenidate (N = 7) [6]. By acquiring multiple hours of baseline, on-drug, and post-drug data from each participant, we showed that acute effects of psilocybin were strongly linked to the subjective psychedelic experience, and changes in hippocampal-cortical connectivity persisted for weeks after. In this study, the effects of methylphenidate in single participants reliably reproduced group effects of stimulants in thousands of individuals in the ABCD study.

Implementation of a PIDT includes longitudinal (rather than cross-sectional) cross-over (rather than parallel groups) design, personalized (rather than group averaged) network mapping and analysis, imaging advances, and careful control for physiological confounds. These advances pivot individual differences from being a limitation of brain imaging science to a strength. This paradigm shift offers hope for expanded utility of fMRI. The PIDT approach could make it possible to conduct target engagement studies that guide key decisions in drug development with fewer participants (Fig. 1). Quantifying the impact of individual variability, measurement reliability, and personalized network mapping on connectivity measurement will allow a complete estimation and optimization of the PIDT approach’s power (Fig. 1, right).

Fig. 1. Power analysis—conventional drug trial designs and precision imaging drug trials.

We estimate the sample size needed to achieve 90% power to detect a difference between drug and placebo on some brain biomarkers, given the constraints listed and a one-sided significance test. Simply switching from a parallel arm to a cross-over design cuts the sample size needed in ¼. PIDT should cut that number down even further (for reasons outlined above), but additional work is needed to accurately assess the benefit. Random intercept STD (standard deviation) measures the similarity of repeated baseline measures. Random slope STD measures similarity of repeated drug exposure effects.

Funding

This work was supported by the Taylor Family Institute Fund for Innovative Psychiatric Research.

Competing interests

JSS is co-author of a provisional patent (Patent 020949/US 15060-1787) for the use of precision functional mapping for measuring target engagement by experimental therapeutics. Within the last year, JSS was an employee of Sumitomo Pharma America and received consulting fees from Longitude Capital. These potential conflicts of interest have been reviewed and are managed by Washington University School of Medicine.