Search bioRxiv⌕ Search
Home/Archive/bioRxiv

bioRxiv · 10.1101/2025.08.29.673066

Shared texture-like representations, not global form, underlie deep neural network alignment with human visual processing

Loke, J. L.·Soerensen, L. K. A.·Groen, I. I. A.·Cappaert, N.·Scholte, H. S.
View original source ↗Export metadata ↓

Abstract

Deep neural networks (DNNs) excel at predicting neural responses across the visual hierarchy 1-5, a success widely interpreted as evidence of shared object recognition computations 6,7. Yet improving DNN object recognition accuracy does not reliably increase neural predictivity 8,9, and even untrained networks predict brain responses above chance 9-11. This disconnect suggests that object recognition may not drive DNN-brain alignment. Texture-like statistics are represented in both DNNs and mid-level visual cortex 12-17. In natural images, these statistics are carried by objects and backgrounds, shaping representations and recognition in both systems 18-21. Does DNN-brain alignment reflect shared sensitivity to object-related information or texture-like statistics? To dissociate these factors, we recorded EEG from 57 participants viewing natural scenes, texture-synthesized images preserving local statistics while disrupting global form and object-only images with backgrounds removed. If alignment reflects texture-like statistics, it should peak for texture-synthesized images. If it reflects object-related processing, alignment should be strongest for natural and object-only conditions, which preserve object information. We compared EEG responses with DNN activations via weighted representational similarity analysis 22,23. Texture-synthesized images yielded the strongest DNN-EEG alignment, peaking in early responses (<200 ms) and explaining up to [~]85% of noise-ceiling-normalized explainable variance versus [~]44% for natural and [~]55% for isolated objects. Crucially, object categories were more decodable for natural and object-only images than texture-synthesized images, yet these object-rich conditions showed weaker alignment. This dissociation reveals that DNNs capture the texture-statistical component of early visual responses while failing to explain later, object-related variance. HighlightsO_LITexture-synthesized images yield the strongest DNN-brain alignment across DNNs. C_LIO_LIObject categories decode better from object-rich than texture-synthesized EEG. C_LIO_LIDNN-brain alignment reflects shared texture statistics, not object-related processing. C_LI

Source connections

Explore related subjects

neurosciencePreprint

Keep this discovery

Explore connections, maps & timelines

BibTeXRIS

Loke, J. L., Soerensen, L. K. A., Groen, I. I. A., Cappaert, N., Scholte, H. S.. 2025-09-04. Shared texture-like representations, not global form, underlie deep neural network alignment with human visual processing. https://doi.org/10.1101/2025.08.29.673066

Cite the original work for its findings. Save a collection to share your selection of sources.

KEEP EXPLORING

Related preprints

bioRxiv2026

The Unreasonable Effectiveness of Cell Types in Describing Neuronal Physiological Features

Single-cell RNA sequencing (scRNA-seq) captures detailed gene expression profiles at scale, while patch-clamp recordings measure intrinsic neuronal electrophysiological properties. Modeling the relations between these two modalities remains a challenge. Here, we compare how well electrophysiological features can be predicted by traditional transcriptomic cell type classification, representations derived from a foundational model (scGPT) pretrained on large-scale scRNA-seq datasets, ion channel-coding genes, and highly variable genes. Using paired transcriptomic and electrophysiological patch-sequencing data from 495 human neurons from neurosurgical tissue, we find that cluster-level cell type representations consistently outperform highly variable gene selection, ion channel gene selection, and context-enriched scGPT embeddings. Notably, performance varies across model architectures and initializations, and the best results are obtained by combining the outputs of separate cell type and scGPT-based models. Together, these findings suggest that traditional discrete cellular classification is highly effective in predicting physiological features. For maximum performance it can be complemented by pretrained transformer models.

neuroscience↗
bioRxiv2026

A nonlinear inhibition pathway underlying cortical responses to tuned holographic optogenetic perturbations

Optogenetics enables causal manipulation of cortical activity. Perturbation responses can be counterintuitive due to network interactions, making theory essential for predicting them. Existing approaches often rely on linear approximations, which fail for many biologically relevant perturbations. Here we develop a nonlinear theory of responses to holographic perturbations in cell-type-specific recurrent networks with structured connectivity. We fit a nonlinear model to mouse V1 data, which shows cotuned-ensemble suppression: perturbing spatially clustered neurons with similar preferred orientations yields markedly stronger short-range suppression than perturbing untuned ensembles. We show that cotuned-ensemble suppression arises from a feature-tuned, nonlinear inhibition pathway implicating somatostatin-positive (SST) interneurons. The theory predicts that cotuned ensembles suppress parvalbumin-positive (PV) neurons but facilitate SST neurons, and links the degree of cotuned-ensemble suppression or facilitation to the variance of the SST response. This framework identifies mechanisms by which nonlinear inhibition sculpts cortical dynamics and establishes a predictive basis for targeted optogenetic interventions.

neuroscience↗
bioRxiv2026

Proteomic signatures of APOE ε4 across human tissues and cell types in Alzheimers disease

The apolipoprotein E {varepsilon}4 (APOE {varepsilon}4) allele is the strongest genetic risk factor for late-onset Alzheimers disease (AD). However, the underlying molecular mechanisms remain unclear. This study included 1691 participants from the Religious Orders Study and Rush Memory and Aging Project (ROSMAP), 1226 participants from the Accelerating Medicines Partnership - Alzheimers Disease (AMP-AD) Diverse Cohorts Study, and 735 participants from the Alzheimers Disease Neuroimaging Initiative (ADNI). To characterise APOE {varepsilon}4 molecular effects, we analysed proteomic data from plasma, cerebrospinal fluid (CSF), and induced pluripotent stem cell (iPSC)-derived astrocytes and neurons, as well as transcriptomic and proteomic data from multiple brain regions. The association of APOE {varepsilon}4 with AD neuropathology was also examined. APOE {varepsilon}4 carriers shared a plasma proteomic signature enriched for immune processes, irrespective of AD diagnosis. A machine learning classifier trained on this signature discriminated APOE {varepsilon}4 carriers from non-carriers in an independent cohort using CSF proteomics. APOE {varepsilon}4 carriage was associated with higher Braak stages and Consortium to Establish a Registry for Alzheimers Disease (CERAD) score. However, only limited APOE {varepsilon}4-associated transcriptomic and proteomic changes were observed in bulk brain tissue, with poor cross-layer concordance. Proteomic analyses of iPSC-derived astrocytes and neurons further revealed cell-type-specific APOE {varepsilon}4-associated changes. APOE {varepsilon}4 is associated with a consistent proteomic signature across plasma and CSF. Its molecular effects in the brain differ across cell types, brain regions and molecular layers. These findings support the need for cell-type-resolved multi-omic studies to elucidate how APOE {varepsilon}4 confers AD risk.

neuroscience↗