Standardizing workflows in imaging transcriptomics with the abagen toolbox

eLife, Journal Year: 2021, Volume and Issue: 10

Published: Nov. 16, 2021

Gene expression fundamentally shapes the structural and functional architecture of human brain. Open-access transcriptomic datasets like Allen Human Brain Atlas provide an unprecedented ability to examine these mechanisms in vivo; however, a lack standardization across research groups has given rise myriad processing pipelines for using data. Here, we develop abagen toolbox, open-access software package working with data, use it how methodological variability influences outcomes Atlas. Applying three prototypical analyses outputs 750,000 unique pipelines, find that choice pipeline large impact on findings, parameters commonly varied literature influencing correlations between derived gene other imaging phenotypes by as much ρ ≥ 1.0. Our results further reveal ordering parameter importance, steps influence normalization yielding greatest downstream statistical inferences conclusions. The presented work development toolbox lay foundation more standardized systematic transcriptomics, will help advance future understanding

Language: Английский

---

Connectome-based modelling of neurodegenerative diseases: towards precision medicine and mechanistic insight

Nature reviews. Neuroscience, Journal Year: 2023, Volume and Issue: 24(10), P. 620 - 639

Published: Aug. 24, 2023

Language: Английский

---

A computational model of neurodegeneration in Alzheimer’s disease

Nature Communications, Journal Year: 2022, Volume and Issue: 13(1)

Published: March 28, 2022

Abstract Disruption of mental functions in Alzheimer’s disease (AD) and related disorders is accompanied by selective degeneration brain regions. These regions comprise large-scale ensembles cells organized into systems for functioning, however the relationship between clinical symptoms dementia, patterns neurodegeneration, functional not clear. Here we present a model association dementia degenerative anatomy using F18-fluorodeoxyglucose PET dimensionality reduction techniques two cohorts patients with AD. This reflected simple information processing-based description macroscale which link to AD physiology, networks, abilities. We further apply normal aging seven diseases functions. propose global processing that links neuroanatomy, cognitive neuroscience neurology.

Language: Английский

---

Large-scale neural dynamics in a shared low-dimensional state space reflect cognitive and attentional dynamics

eLife, Journal Year: 2023, Volume and Issue: 12

Published: July 3, 2023

Cognition and attention arise from the adaptive coordination of neural systems in response to external internal demands. The low-dimensional latent subspace that underlies large-scale dynamics relationships these cognitive attentional states, however, are unknown. We conducted functional magnetic resonance imaging as human participants performed tasks, watched comedy sitcom episodes an educational documentary, rested. Whole-brain traversed a common set states spanned canonical gradients brain organization, with global desynchronization among networks modulating state transitions. Neural were synchronized across people during engaging movie watching aligned narrative event structures. reflected fluctuations such different indicated engaged task naturalistic contexts, whereas lapses both contexts. Together, results demonstrate traversals along organization reflect dynamics.

Language: Английский

---

Representation learning of resting state fMRI with variational autoencoder

NeuroImage, Journal Year: 2021, Volume and Issue: 241, P. 118423 - 118423

Published: July 23, 2021

Resting state functional magnetic resonance imaging (rsfMRI) data exhibits complex but structured patterns. However, the underlying origins are unclear and entangled in rsfMRI data. Here we establish a variational auto-encoder, as generative model trainable with unsupervised learning, to disentangle unknown sources of activity. After being trained large from Human Connectome Project, has learned represent generate patterns cortical activity connectivity using latent variables. The representation its trajectory spatiotemporal characteristics variables reflect principal gradients drive changes networks. Representational geometry captured covariance or correlation between variables, rather than connectivity, can be used more reliable feature accurately identify subjects group, even if only short period is available each subject. Our results demonstrate that VAE valuable addition existing tools, particularly suited for learning resting fMRI

Language: Английский

---

Deciphering the clinico-radiological heterogeneity of dysexecutive Alzheimer’s disease

Cerebral Cortex, Journal Year: 2023, Volume and Issue: 33(11), P. 7026 - 7043

Published: Jan. 31, 2023

Dysexecutive Alzheimer's disease (dAD) manifests as a progressive dysexecutive syndrome without prominent behavioral features, and previous studies suggest clinico-radiological heterogeneity within this syndrome. We uncovered using unsupervised machine learning in 52 dAD patients with multimodal imaging cognitive data. A spectral decomposition of covariance between FDG-PET images yielded six latent factors ("eigenbrains") accounting for 48% variance patterns hypometabolism. These eigenbrains differentially related to age at onset, clinical severity, performance. hierarchical clustering on the eigenvalues these four subtypes, i.e. "left-dominant," "right-dominant," "bi-parietal-dominant," "heteromodal-diffuse." Patterns hypometabolism overlapped those tau-PET distribution MRI neurodegeneration each subtype, whereas amyloid deposition were similar across subtypes. Subtypes differed onset severity where heteromodal-diffuse exhibited worse picture, bi-parietal had milder presentation. propose conceptual framework executive components based associations observed dAD. demonstrate that dAD, despite sharing core are diagnosed variability their neuroimaging profiles. Our findings support use data-driven approaches delineate brain-behavior relationships relevant practice physiology.

Language: Английский

---

Test–retest reliability and predictive utility of a macroscale principal functional connectivity gradient

Human Brain Mapping, Journal Year: 2023, Volume and Issue: 44(18), P. 6399 - 6417

Published: Oct. 18, 2023

Abstract Mapping individual differences in brain function has been hampered by poor reliability as well limited interpretability. Leveraging patterns of brain‐wide functional connectivity (FC) offers some promise this endeavor. In particular, a macroscale principal FC gradient that recapitulates hierarchical organization spanning molecular, cellular, and circuit level features along sensory‐to‐association cortical axis emerged both parsimonious interpretable measure behavior. However, the measurement reliabilities have not fully evaluated. Here, we assess global regional measures using test–retest data from young adult Human Connectome Project (HCP‐YA) Dunedin Study. Analyses revealed were (1) consistently higher than those for traditional edge‐wise measures, (2) derived general (GFC) comparison with resting‐state FC, (3) longer scan lengths. We additionally examined relative utility these predicting cognition aging datasets HCP‐aging dataset. These analyses range significantly associated all three datasets, moderately HCP‐YA Study reflecting contractions expansions hierarchy, respectively. Collectively, results demonstrate gradient, especially GFC, effectively capture reliable feature human subject to biologically meaningful variation, offering advantages over search brain–behavior associations.

Language: Английский

---

Omnipresence of the sensorimotor-association axis topography in the human connectome

NeuroImage, Journal Year: 2023, Volume and Issue: 272, P. 120059 - 120059

Published: March 30, 2023

Low-dimensional representations are increasingly used to study meaningful organizational principles within the human brain. Most notably, sensorimotor-association axis consistently explains most variance in connectome as its so-called principal gradient, suggesting that it represents a fundamental principle. While recent work indicates these low dimensional relatively robust, they limited by modeling only certain aspects of functional connectivity structure. To date, majority studies have restricted approaches strongest connections brain, treating weaker or negative noise despite evidence structure among them. The present examines gradients across full range strengths and explores implications for outcomes individual differences, identifying potential dependencies on thresholds opportunities improve prediction tasks. Interestingly, emerged gradient entire levels. Moreover, at intermediate encoded better followed individual-specific anatomical features, was also more predictive intelligence. Taken together, our results add principle brain's organization, since is evident even lenient thresholds. These loosely coupled further appear contain valuable potentially important information could be understanding diagnosis, treatment outcomes.

Language: Английский

---

Toward a more informative representation of the fetal–neonatal brain connectome using variational autoencoder

eLife, Journal Year: 2023, Volume and Issue: 12

Published: May 15, 2023

Recent advances in functional magnetic resonance imaging (fMRI) have helped elucidate previously inaccessible trajectories of early-life prenatal and neonatal brain development. To date, the interpretation fetal–neonatal fMRI data has relied on linear analytic models, akin to adult neuroimaging data. However, unlike brain, fetal newborn develops extraordinarily rapidly, far outpacing any other development period across life span. Consequently, conventional computational models may not adequately capture these accelerated complex neurodevelopmental during this critical along prenatal-neonatal continuum. obtain a nuanced understanding development, including nonlinear growth, for first time, we developed quantitative, systems-wide representations activity large sample (>500) fetuses, preterm, full-term neonates using an unsupervised deep generative model called variational autoencoder (VAE), shown be superior representing resting-state healthy adults. Here, demonstrated that features, is, latent variables, derived with VAE pretrained rsfMRI human adults, carried important individual neural signatures, leading improved representation maturational patterns more accurate stable age prediction neonate cohort compared models. Using decoder, also revealed distinct networks spanning sensory default mode networks. VAE, are able reliably quantify complex, connectivity. This will lay foundation detailed mapping aberrant signatures their origins life.

Language: Английский

---

Fronto-occipital dyscommunication associates with brain hierarchy in schizophrenia

Communications Biology, Journal Year: 2025, Volume and Issue: 8(1)

Published: May 5, 2025

Language: Английский

---