Editorial: Emerging trends in large-scale data analysis for neuroscience research

The primary aim of this research topic is to showcase recent progress in data-driven approaches for studying the brain. It focuses on tackling challenges in managing, processing, and interpreting large-scale neuroscience data while identifying future research opportunities. This topic will delve into state-of-the-art tools and methods for analyzing, integrating, and interpreting extensive neuroscience datasets.Excluding the retraction, there are five papers published in this research topic. Hsu et al. consider the problem of warping and registering brain images to a standard template, which can introduce spatial errors and reduce accuracy. They develop LYNSU (Locating by YOLO and Segmenting by U-Net), an automated method for segmenting neuropils in fluorescence images from the FlyCircuit database, eliminating the need for warping and facilitating high-throughput anatomical analysis and connectomics in the Drosophila brain. They demonstrate performance comparable to manual annotations, with a 3D Intersection-over-Union (IoU) of 0.869, and segments a neuropil in about 7 seconds.Miranda considers task-based fMRI studies and develops a fast Bayesian function-onscalar model to estimate population-level activation maps for the working memory task. The proposed approach uses a canonical polyadic (CP) tensor decomposition to extract shared and subject-specific features from individual coeBicient maps. The subject-specific features are modeled as functions of covariates within a Bayesian framework that accounts for correlations in the CP-extracted features. The proposed decomposition facilitates fast computation and allows eBicient MCMC estimation of population-level activation maps.Dang, Fermin and Machizawa consider the problem of decoding and feature selection in high dimensions. They introduce the optimized Forward Variable Selection Decoder (oFVSD) toolbox as a feature selection methodology that combines forward variable selection (FVS) and hyperparameter optimization integrated with 18 machine learning models. They test sex classification and age range regression on 1,113 structural MRI datasets and demonstrate performance improvements over models without FVS. The methodology is available as an open-source Python package. Bologna et al. consider the construction of data-driven brain models using neural simulation environments and large-scale computing facilities. They developed the EBRAINS Hodgkin-Huxley Neuron Builder (HHNB), a web resource for building single cell neural models via the extraction of activity features from electrophysiological data with estimation based on a genetic algorithm. HHNB then allows simulation of the brain model using the estimated model through an interactive setting. Kim et al. consider the visualization of gene expression obtained using RNA sequencing across the brain. Molecular patterns emerging from spatial transcriptomic data can be associated with circuitry and function in the neocortex. They propose a web app LaminaRGeneVis for visualizing laminar gene expression across datasets collected using bulk, single-nucleus, and spatial RNA sequencing. Allowing for normalizations across diBerent datasets, the app supports single-and multi-gene analyses, data visualization and statistics for the adult human neocortex.

Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.