Reliability and predictability of phenotype information from functional connectivity in large imaging datasets

Jessica Dafflon,Dustin Moraczewski,Eric Earl,Dylan M. Nielson,Gabriel Loewinger,Patrick McClure,Adam G. Thomas,Francisco Pereira

2024-05-01

Abstract:One of the central objectives of contemporary neuroimaging research is to create predictive models that can disentangle the connection between patterns of functional connectivity across the entire brain and various behavioral traits. Previous studies have shown that models trained to predict behavioral features from the individual's functional connectivity have modest to poor performance. In this study, we trained models that predict observable individual traits (phenotypes) and their corresponding singular value decomposition (SVD) representations - herein referred to as latent phenotypes from resting state functional connectivity. For this task, we predicted phenotypes in two large neuroimaging datasets: the Human Connectome Project (HCP) and the Philadelphia Neurodevelopmental Cohort (PNC). We illustrate the importance of regressing out confounds, which could significantly influence phenotype prediction. Our findings reveal that both phenotypes and their corresponding latent phenotypes yield similar predictive performance. Interestingly, only the first five latent phenotypes were reliably identified, and using just these reliable phenotypes for predicting phenotypes yielded a similar performance to using all latent phenotypes. This suggests that the predictable information is present in the first latent phenotypes, allowing the remainder to be filtered out without any harm in performance. This study sheds light on the intricate relationship between functional connectivity and the predictability and reliability of phenotypic information, with potential implications for enhancing predictive modeling in the realm of neuroimaging research.

Neurons and Cognition

What problem does this paper attempt to address?

The main problem that this paper attempts to solve is to improve the reliability and accuracy of predicting phenotypic information from brain functional connectivity data. Specifically, the researchers are concerned with how to predict an individual's behavioral characteristics (i.e., phenotypes) through the functional connectivity patterns in large - scale imaging datasets. Although previous studies have shown that models trained from an individual's functional connectivity have relatively limited performance in predicting behavioral characteristics, this paper explores the predictive performance of these models by using two different large - scale neuroimaging datasets - the Human Connectome Project (HCP) and the Philadelphia Neurodevelopmental Cohort (PNC), and by applying Singular Value Decomposition (SVD) to represent latent phenotypes.

The main objectives of the paper include:
1. **Evaluating the predictive performance of latent phenotypes**: The researchers compared the performance of directly predicting the original phenotypes from the functional connectivity data with predicting their corresponding SVD representations (i.e., latent phenotypes).
2. **Exploring the importance of the first few latent phenotypes**: The study found that only the first 5 latent phenotypes could be reliably identified, and the prediction using these 5 reliable latent phenotypes was similar to the prediction performance using all latent phenotypes.
3. **Analyzing the factors influencing the predictive performance**: The study also explored the effect of removing confounding factors (such as age and sex) on the predictive performance, and how the reliability of latent phenotypes affects their predictability from the functional connectivity data.

Through these studies, the paper aims to reveal the complex relationship between functional connectivity and phenotypic information prediction, and to provide directions for improvement in predictive modeling in future neuroimaging studies.