Generalizable Machine Learning in Neuroscience Using Graph Neural Networks

Abstract

Although a number of studies have explored deep learning in neuroscience, the application of these algorithms to neural systems on a microscopic scale, i.e. parameters relevant to lower scales of organization, remains relatively novel. Motivated by advances in whole-brain imaging, we examined the performance of deep learning models on microscopic neural dynamics and resulting emergent behaviors using calcium imaging data from the nematode C. elegans. As one of the only species for which neuron-level dynamics can be recorded, C. elegans serves as the ideal organism for designing and testing models bridging recent advances in deep learning and established concepts in neuroscience. We show that neural networks perform remarkably well on both neuron-level dynamics prediction and behavioral state classification. In addition, we compared the performance of structure agnostic neural networks and graph neural networks to investigate if graph structure can be exploited as a favourable inductive bias. To perform this experiment, we designed a graph neural network which explicitly infers relations between neurons from neural activity and leverages the inferred graph structure during computations. In our experiments, we found that graph neural networks generally outperformed structure agnostic models and excel in generalization on unseen organisms, implying a potential path to generalizable machine learning in neuroscience.

Connecting the dots between cell populations, whole-brain activity, and behavior

“…A few groups are attempting to find solutions to help with data management, including the use of more automated methods, such as ICA 148 and machine learning techniques. 149 Choosing the most scientifically sound way to process and analyze concurrent fMRI and optogenetic data will continue to be an important consideration in studies applying these techniques.…”

Learning to Model the Relationship Between Brain Structural and Functional Connectomes

“…Indeed, the GRL paradigm is to learn low-dimensional embeddings of individual vertices, edges, or the graph itself [23], [32]- [34], which can then be used in e.g., (semi-supervised) node classification [15], link prediction [35], graph clustering [36], [37], and graph clas-sification [38]. Recently, GRL ideas have permeated to neuroimaging data analysis for behavioral state classification [39], to study the relationship between SC and FC [13], [40], and to extract representations for subject classification [41]- [43]. Although these prior works have probed the area of brain connectomics in several novel directions and achieved solid performance in multiple regression or classification tasks, they mostly rely on a single type of brain network and use feedforward models to predict subject labels from input graphs.…”

Predicting in vitro single-neuron firing rates upon pharmacological perturbation using Graph Neural Networks

“…In neuroscience, GNNs have recently been shown to be effective in several tasks, such as classification of brain states (Bessadok et al, 2019 ; Banka et al, 2020 ; Lostar and Rekik, 2020 ; Cui et al, 2021 ; Li et al, 2021 ; Wein et al, 2021 ; Xing et al, 2021 ), detection of the default mode network (Wang et al, 2022 ), brain parcellation (Eschenburg et al, 2021 ; Qiu et al, 2022 ), and disease detection (Chen et al, 2021 ; Chan et al, 2022 ) based on functional connectivity derived from functional magnetic resonance imaging data. At the neuron level, GNNs were used to model motor action trajectories in C. elegans using connectivity graphs derived from calcium imaging of individual neurons (Wang et al, 2021 ).…”