Toward a Connectivity Gradient-Based Framework for Reproducible Biomarker Discovery

Abstract

Despite myriad demonstrations of feasibility, the high dimensionality of fMRI data remains a critical barrier to its utility for reproducible biomarker discovery. Recent studies applying dimensionality reduction techniques to resting-state fMRI (R-fMRI) have unveiled neurocognitively meaningful connectivity gradients that are present in both human and primate brains, and appear to differ meaningfully among individuals and clinical populations. Here, we provide a critical assessment of the suitability of connectivity gradients for biomarker discovery.Using the Human Connectome Project (discovery subsample=209; two replication subsamples= 209×2) and the Midnight scan club (n=9), we tested the following key biomarker traits -reliability, reproducibility and predictive validity -of functional gradients. In doing so, we systematically assessed the effects of three analytical settings, including i) dimensionality reduction algorithms (i.e., linear vs. non-linear methods), ii) input data types (i.e., raw time series, [un-]thresholded functional connectivity), and iii) amount of the data (R-fMRI time-series lengths). We found that the reproducibility of functional gradients across algorithms and subsamples is generally higher for those explaining more variances of whole-brain connectivity data, as well as those having higher reliability. Notably, among different analytical settings, a linear dimensionality reduction (principal component analysis in our study), more conservatively thresholded functional connectivity (e.g., 95-97%) and longer time-series data (at least ≥20mins) was found to be preferential conditions to obtain higher reliability. Those gradients with higher reliability were able to predict unseen phenotypic scores with a higher accuracy, highlighting reliability as a critical prerequisite for validity. Importantly, prediction accuracy with connectivity gradients exceeded that observed with more traditional edge-based connectivity measures, suggesting the added value of a low-dimensional gradient approach. Finally, the present work highlights the importance and benefits of systematically exploring the parameter space for new imaging methods before widespread deployment.

Visual statistical learning alters low-dimensional cortical architecture

“…Next, to investigate changes in cortical and subcortical patterns of functional connectivity during SL, we used the centered matrices to estimate separate cortical-subcortical connectivity manifolds for each participant's Pre-learning, Early-learning, Late-learning, and Post-learning covariance matrices. Building from prior work (36,(59)(60), we converted each centered connectivity matrix into an affinity matrix before applying PCA to the entire sample's affinity matrices. This reveals a set of principal components (PCs), which offer a low-dimensional representation of the cortical and subcortical structure (i.e., a cortical-subcortical manifold).…”

Functional and microstructural plasticity following social and interoceptive mental training

“…The dimensionality reduction techniques applied here are particularly suitable, as they provide an intrinsic and continuous coordinate system that situates individual regions within the global landscape of the cortical connectome. Gradients in our sample of middle old healthy individuals who underwent extensive screening before study enrollment (Singer et al, 2016), recapitulated prior work in healthy young adults (Hong et al, 2020;Margulies et al, 2016). Indeed, in accordance to prior studies, we observed that G1 traversed a sensory-fugal axis, running from unimodal to transmodal systems such as the default mode and fronto-parietal networks.…”

“…Such large-scale functional gradients can be reliably identified from in vivo functional connectivity information, by applying techniques that identify principal dimensions of connectivity variations from one region to the next. As such, they provide a low-dimensional coordinate system reflecting macroscale neural organization, specifically accounting for both functional segregation across regions as well as their integration (Haak and Beckmann, 2020;Haak et al, 2018;Hong et al, 2020;Margulies et al, 2016;Shine et al, 2019a).…”

fMRI connectivity as a biomarker of antipsychotic treatment response: A systematic review

“…Concerning fMRI acquisition, previous studies have shown that many non-neural factors may influence the variation in fMRI signals and, consequently, the reliability of the intra- and inter-individual FC ( Hong et al, 2020 , Birn et al, 2008 ). One example is scanning duration.…”

“…For example, four studies used global signal regression for physiological motion correction, which is a risk of discarding global neural information ( Power et al, 2014 , Murphy and Fox, 2017 ). Recent work explored a set of benchmark parameters for the use of FC as a tool in the domain of biomarker discovery including: using a linear dimensionality reduction algorithm, utilizing the gradients that explain a greater amount of the variance of the data and extracting gradients using more conservatively threshold FC matrices ( Hong et al, 2020 ). Future work in this direction and exploration of large-sample test–retest neuroimaging datasets, will be valuable in biomarker development and may impact the interpretation of findings summarized in this review.…”