Multimodal Brain Data Integration

25.04.2026, Saturday, 12:00-13:30

General focus of the symposium:

Modeling and understanding the brain, its development, function, and evolution, present several challenges. To fully comprehend the intricate dynamics of neural architecture across spatial and temporal scales, multimodal integration ranging from neurogenetics to developmental dynamics and functional organization is required. Recent advancements in genetic profiling, brain atlas learning, connectome analysis, and cutting-edge neuroimaging technologies have facilitated multi-scale investigations of brain networks and morphology. This symposium emphasizescomputational strategies for integrating such diverse data sources into multimodal brain atlases. It explores how to mine such heterogeneous brain datasets for functional insights across different model organisms, such as fruit flies (Drosophila melanogaster) and mice, as well as human data. This approach can reveal complex organizational principles that link neurogenetics to functional connectivity and network graphs, as well as trace patterns of neurocognitive evolution across species in silico. Such computational methods are increasingly vital because they leverage the expanding universe of multimodal brain data for resource-efficient hypothesis generation, whether for more refined experimental testing or in experimentally inaccessible cases.

In particular, we want to show how the tool BrainTrawler provides a low-threshold to a multimodal database consisting of transcriptomic, connectivity, and activity data to explore human and mouse brain space. Furthermore, we highlight how the LarvalBrain tool offers unprecedented access and query possibilities for unique data collectionof fruit fly larvae across different developmental stages, enabling novel workflows that support behavioral experiments. Furthermore, we present a mouse case study demonstrating how the integration of multiparametric MRI with gene expression data can improve the interpretation of neuroimaging results and aid in identifying brain regions engaged by pharmacological interventions. Finally, we showcase how fusing paleogenomics with publicly available brain data allows for the exploration of the cognitive and socio-affective trait space of archaic brains for the computational reconstruction of brain evolution.

In summary, we aim to demonstrate and raise awareness of complementary approaches that computationally mine the ever-expanding public multimodal brain databases for the much-needed integrated exploration and understanding of brain function, morphology, and evolution across biological levels and model organisms.

12:00 Sophia Ulonska, Ph.D

Biomedical Image Informatic Group, VRVis GmbH, Vienna, Austria

"BrainTrawler: An Interactive Platform for Exploring Multimodal Data in Humans and Mice"

Sophia Ulonska (VRVis GmbH) will showcase how multimodal brain data are integrated into the data resource Brain Transcriptomic and Connectivity Data (BrainTACO) within the free web-based platform BrainTrawler. She will highlight an innovative and unique method for deriving circuits (neural networks) from brain activity signals in real time, with the capability to mine differentially expressed genes in the entire circuit or its components. This enables multimodal exploration, from single-cell gene expression to connectivity and brain activity across brain initiatives and species in a 3D context in both humans and mice. BrainTrawler aims to support researchers in accelerating data-driven insights and hypothesis generation to understand the brain. In addition, it supports the teaching of brain-gene-function relationships. Furthermore, it is valuable to identify potential target genes in the context of meso-scale brain circuits.

12:20 Tobias Peherstorfer

Biomedical Image Informatic Group, VRVis GmbH, Vienna, Austria

"LarvalBrain: A Multimodal Drosophila Resource for Genetic Tool Discovery"

Tobias Peherstorfer (VRVis GmbH) will demonstrate how data across different developmental stages of the fruit fly (Drosophila melanogaster) larva can be integrated into a standardized atlas framework. By leveraging a shared spatial context, integrated data can be mined in an anatomical and genetic context. He will demonstrate this approach in the web-based data exploration tool LarvalBrain, showcasing how light and electron microscopy imaging data from different larval instars can be explored and searched interactively to inform future behavioral experiments. This software enables researchers to easily access complex and multimodal neuroscientific data without requiring programming skills. In particular, he will highlight a novel approach for discovering potential split-line parents by mining large Gal4 light microscopy image collections from the 3rd larval instar. Additionally, he will demonstrate an approach to link high-resolution electron microscopy neuron tracing to Gal4 staining data.

12:40 Lukasz Piszczek, Ph.D

Medical University of Vienna, Austria

"Computational Reconstruction of Evolutionary Selection Shaping Socio-Affective Traits in Human Brain Networks"

Dr. Lukasz Piszczek will address the ongoing challenges in understanding human brain evolution. He will specifically highlight an integrated strategy that offers a holistic workflow for utilizing publicly available ancient genomic and multimodal brain data in the context of computational reconstruction of brain evolution. This approach aims to map multigenic evolution across cognitive, cellular, and molecular levels, thereby exploring cognitive dynamics and behavioral traits over evolutionary scales. Such a workflow can provide a valuable computational framework at the mesoscale level, acting as a crucial link between molecular/cellular (bottom-up) and system (top-down) level data, thus facilitating a comprehensive computational archaeology of the human brain. By mining an expanded dataset of human task functional networks, it becomes possible to infer the functional features of archaic brains from extinct hominin genomes and extrapolate the associated cognitive attributes and socio-affective traits.

13:00 Natalia K. Freus

Institute for Experimental Pharmacology and Toxicology, Friedrich-Alexander-University Erlangen-Nuremberg, Germany

"Pharmacological and Resting State fMRI Reveal Osteocalcin's Effects on Mouse Brain Regions with High Gpr37 and Gpr158 Expression"

Natalia Freus will present how multiparametric magnetic resonance imaging (MRI) can be combined with gene expression data to identify brain regions influenced by the bone-derived hormone osteocalcin (OCN). OCN is a key component of the bone-brain axis and signals via the Gpr158 and Gpr37 receptors in the mouse brain. The absence of OCN in knockout mice has been linked to increased anxiety and depression-like behaviors. Using resting-state functional MRI (rs-fMRI) and relative cerebral blood volume (rCBV) mapping in mice after systemic OCN administration, the directly responsive and secondarily modulated brain regions were characterized. Gene expression maps from the Allen Brain Institute were used to highlight regions with high OCN receptor expression and to relate receptor distribution to functional effects. This integrative approach demonstrates how combining rs-fMRI, rCBV, and gene expression data provides a powerful strategy for inferring the central mode of action of neuromodulatory hormones or candidate drugs.

13:20 Julian Kędys

Poznań Supercomputing and Networking Center, Polish Academy of Science

"Consensus Shared Alignment (CoSA): Fair and Robust Alignment of Subjects with Heterogeneous Brain Activity Patterns through a Multi-Method Meta-Approach"

Studying neurodevelopmental conditions of complex clinical phenotypes is among the most critical and simultaneously the most challenging endeavours in modern neuroscience. Population-level alignment and dimensionality reduction of brain-activity time series enable cross-subject comparisons - essential for productive, rigorous research in the domain - but a single alignment method can favour subjects whose dynamics match its inductive bias, diluting subgroup-specific temporal structure in the shared space.

We construct a controllable, transparent meta-alignment that fuses complementary population methods to obtain a shared latent representation that is robust and comparatively fair across heterogeneous cohorts.

We propose Consensus Shared Alignment (CoSA). For each method (e.g., group principal component analysis (PCA), independent component analysis (ICA), shared response model (SRM), multi-set canonical correlation analysis (MCCA)), we estimate low-dimensional loadings and align them to a common reference basis using an orthogonal Procrustes map. Components are matched one-to-one with a Hungarian algorithm and signs are harmonised. Aligned loadings are fused with non-negative weights that sum to one, chosen from cross-method agreement and stability diagnostics. Subject latent trajectories are then computed and standardised within subject. CoSA reports permutation-based reproducibility, temporal continuity and subsample-stability checks.

On whole-brain rodent imaging data (controls + autism-model subtypes), CoSA yielded stable shared coordinates without single-method dominance. Relative to individual methods, CoSA increased cross-subject reproducibility while preserving subtype-specific temporal structure, enabling direct subgroup-faithful comparisons.

Multi-method consensus alignment can reduce inductive-bias effects in population embeddings and provides practical diagnostics for reliability and control, supporting fairer analyses in heterogeneous neuroimaging data.