Biology subjects

Liu, T.

Publications and source records attributed to Liu, T..

At least 109 records · Page 6Linked to original sources

bioRxiv2025

Convergent utilization of APN receptor by hedgehog merbecoviruses

Dipeptidyl peptidase-4 (DPP4) and angiotensin-converting enzyme 2 (ACE2) are well-established receptors for merbecoviruses, yet the receptor usage of merbecoviruses of European and Asian hedgehogs (EriCoVs) remains unknown. Here, by testing hedgehog orthologs of known coronavirus receptors, we identify hedgehog aminopeptidase N (APN) as a functional receptor for EriCoVs. Analysis of 94 APN orthologs indicates that EriCoVs have a limited host range, primarily utilizing hedgehog APN and, to a lesser extent, APN from certain felids, shaped by specific determinants at the virus-receptor interface. Cryo-EM reveals an APN-binding mode distinct from those used by alpha- and deltacoronaviruses. Functional assays indicate that hedgehog transmembrane serine protease 2 (TMPRSS2) enhances spike activation and promotes pseudovirus entry. Neutralizing antibodies targeting RBD and APN were developed and could effectively block EriCoV pseudovirus entry and propagation. These findings reveal an unexpected convergent evolution of APN utilization among merbecovirus, establishing a foundation for risk assessment and countermeasure development. Graphic abstractConvergent utilization of hedgehog APN by EriCoVs O_FIG O_LINKSMALLFIG WIDTH=200 HEIGHT=115 SRC="FIGDIR/small/676546v1_ufig1.gif" ALT="Figure 1"> View larger version (48K): org.highwire.dtl.DTLVardef@1ea334aorg.highwire.dtl.DTLVardef@cd53bforg.highwire.dtl.DTLVardef@352102org.highwire.dtl.DTLVardef@a1eed_HPS_FORMAT_FIGEXP M_FIG C_FIG

molecular biology↗

bioRxiv2025

Resting-State fMRI and the Risk of Overinterpretation: Noise, Mechanisms, and a Missing Rosetta Stone

Resting-state fMRI has generated influential insights into large-scale brain organization and contributed to clinically relevant applications, largely through correlation-based measures of cross-regional association in BOLD responses. At the same time, interpreting these statistical associations as reflecting underlying neural interactions requires careful consideration of fundamental methodological constraints. Here, we distinguish two fundamental but often conflated limitations. The first is measurement distortion: the fMRI signal is an indirect and heterogeneous measurement of neural activity, arising from neurovascular coupling, physiology, and measurement-related processes, which can introduce systematic and incompletely characterized biases into estimated correlations. The second is causal non-identifiability: even if correlations perfectly reflected neural synchrony, the resulting correlation structure would not uniquely determine the underlying neural interactions. Using causal reasoning, simulations, and analytic arguments, we distinguish these limitations and examine their consequences for interpretation. Although both apply broadly to fMRI, their implications are particularly important in resting-state analyses, where correlation structure is the primary object of inference in the absence of experimental perturbation. We show that measurement-related biases can distort estimated correlations (e.g., attenuating or in-flating associations and affecting group comparisons) and can produce reproducible patterns that do not necessarily reflect underlying neural relationships, highlighting that statistical reliability does not guarantee biological validity. We further show that graph-theoretic, geometric, and other higher-order representations derived from these correlations do not, by themselves, justify mechanistic interpretation. We argue not against the utility of resting-state fMRI, but for greater precision in interpretation. Our conclusions concern the interpretation of correlation-based analyses rather than their methodological utility. Distinguishing measurement distortion from causal non-identifiability clarifies the inferential boundaries separating descriptive association, predictive utility, and mechanistic interpretation.

neuroscience↗

bioRxiv2025

Fluorescent protein-based ticker tapes for multiplexed recordings of transcriptional histories in single cells in culture and in vivo

Recording and imaging of promoter activities in real time is critical for deciphering cellular states and dynamic signaling crosstalk, but technologies capable of simultaneously capturing multiple transient events in living cells are lacking. Here, we custom-design fluorescent protein-based ticker tapes (FPTT) for multiplexed and longitudinal recording of physiological activities in single cells. FPTT integrates multi-spectral monomeric fluorescent proteins with self-assembling protein fibers, enabling massively parallel analysis of signaling dynamics under varying cellular conditions. Using FPTT, we were able to log dose-dependent and reversible transcriptional histories of endogenous cFos signaling in primary hippocampus neurons at a 3-hour temporal resolution via biological timestamps and an extended recording time of over 8 days. Furthermore, we expanded the imaging toolset by engineering genetically encoded FPTT variants for human NF{kappa}B, JAK/STAT3, mTOR, NFAT and cAMP signaling, allowing for the quantification of potential crosstalk between cFos and NF{kappa}B pathways in neurons, multiplexed recording of STAT3- and cAMP-specific promoters during drug-induced liver injury in mice, as well as simultaneous analysis of up to four major signaling pathways involved in experimental T-cell activation. This platform advances single-cell analysis by providing a versatile tool to investigate transcriptional histories and signaling interplay across diverse biological contexts, with broad applications in developmental biology and disease modeling.

synthetic biology↗

bioRxiv2025

Highly controllable co-delivery of siRNA and doxorubicin via cationic niosomes for synergistic anticancer effects

The combination of anticancer drugs and siRNA has emerged as a promising strategy for cancer therapy due to their synergistic effects. However, a major challenge is developing a co-delivery system with high controllability capable of simultaneously encapsulating both the chemical drug and siRNA. In this study, we developed a simple and highly controllable co-delivery system for delivering Doxorubicin (Dox) and siRNA. Co-encapsulation was achieved by intercalating Dox into duplex GC-rich oligodeoxynucleotides, (CGA)7/(TCG)7 (termed CGA-Dox), followed by mixing with siRNA and subsequent condensation using a cationic noisome (SPN) through electrostatic interactions. SPN/siRNA/Dox exhibited efficient cellular uptake and high gene silencing efficiency in both H1299 and MDA-MB-231 cells. Furthermore, co-delivery of Dox and siRNA targeting ribonucleotide reductase subunit M2 (RRM2) using this system significantly enhanced cellular apoptosis and antiproliferative effect in MDA-MB-231 cells compared to Dox alone. These results indicate that our co-delivery system constructed via Dox-DNA intercalation and electrostatic complexation, represents a promising strategy for synergistic anticancer therapy.

bioengineering↗

bioRxiv2025

Metabolic Immunosuppression Mediated by Proliferative CD168+TAMs in Colon Cancer

The role of non-monocyte-derived, in situ proliferating tumor associated macrophages (TAMs) in tumor progression has been underappreciated. Proliferative TAM subsets likely represent critical mediators in establishing the immunosuppressive microenvironment, given the susceptibility of TAMs to be hijacked by malignant cells for pro-tumorigenic functions. However, the regulatory networks governing their self-renewal circuitry and lineage bifurcation remained undefined. Combining single cell RNA sequencing of human and murine colon tumors, we identified CD168+ TAMs, a hyperproliferative subpopulation with FOXM1-dependent maintenance of replicative competence. Spatial multi-omics and functional validation revealed that this subset exhibited peritumoral localization and underwent immunosuppressive polarization through cholesterol-coordinated CX3CL1-CX3CR1 signaling. Colon cancer-specific lipid metabolic rewiring elevates glycerophospholipid and cholesterol flux, with PLA2G7-mediated glycerophospholipid catabolism promoting CX3CR1 expression in TAMs. Thus, the CX3CR1 signaling-driven immunosuppressive effect of CD168+ TAM is an integrated response to tumor-specific metabolic features. Targeting the CD168+ TAMs may enable therapeutic strategies for metabolic-immune reshaping in colon cancers.

cancer biology↗

bioRxiv2025

Lifestyle and transcriptional signatures associated with ethnicity/race-related variations in the functional connectome

Understanding the variation of functional architecture across individuals and populations is fundamental to advancing our knowledge of human health and behaviour. Yet, while functional organization differences related to ethnicity/race are consistently reported, their underlying mechanisms remain poorly understood. Here, we apply precision individualized functional mapping to systematically investigate ethnicity/race-related differences in the brains intrinsic organization and their associations with lifestyle and transcriptional signatures. We show that variations in network topography and functional connectivity across ethnic/racial groups follow a hierarchical sensorimotor- association axis and are constrained by brain morphology. Importantly, we identify lifestyle factors-- particularly education and substance use--that significantly mediate these associations between ethnicity/race and functional connectivity. Leveraging human brain gene expression data, we further demonstrate that cortical transcriptional patterns are spatially aligned with ethnicity/race-related variability in functional connectivity. Gene ontology analyses of associated genes reveal significant enrichment in biological processes such as synaptic signalling and neuronal system development. Together, these findings uncover a multi-layered framework linking ethnicity/race-related differences in brain function to structural constraints, lifestyle influences, and molecular signatures, and advance a more comprehensive and equitable understanding of human brain diversity.

neuroscience↗

bioRxiv2025

Structural basis of NSD2 degradation via targeted recruitment of SCF-FBXO22

Targeted protein degradation (TPD) through the ubiquitin-proteasome system is driven by compound-mediated polyubiquitination of a protein-of-interest by an E3 ubiquitin (Ub) ligase. To date, relatively few E3s have been successfully utilized for TPD and the governing principles of functional ternary complex formation between the E3, degrader, and protein target remain elusive. FBXO22 has recently been harnessed by several groups to target different proteins for degradation. FBXO22 recruitment has been enabled through degraders that covalently modify its cysteine residues. Here, we reveal that the aldehyde derivative of UNC10088 promotes cooperative binding of FBXO22 to NSD2, a histone methyltransferase and oncogenic protein, leading to a cryo-EM structure of the full SKP1-CUL1-F-box (SCF)-FBXO22 complex with NSD2. This structure revealed a conformational change in the FBXO22 loop surrounding C326, further exposing the cysteine for covalent recruitment. Additional medicinal chemistry efforts led to the discovery of benzaldehyde-based non-prodrug degraders that similarly engage C326 of FBXO22 and potently degrade NSD2. Furthermore, unlike many degraders, our molecules recruit NSD2 to a different surface of FBXO22 than the known FBXO22 substrate BACH1, allowing for concurrent complex formation and degradation of both the neosubstrate and endogenous substrates. Overall, we demonstrate the biochemical and structural basis for NSD2 degradation, revealing key principles for efficient and selective TPD by SCFFBXO22.

biochemistry↗

bioRxiv2025

Transcriptome Universal Single-isoform COntrol: A Framework for Evaluating Transcriptome reconstruction Quality

Long-read sequencing (LRS) platforms, such as Oxford Nanopore and Pacific Biosciences, enable comprehensive transcriptome analysis but face challenges such as sequencing errors, sample quality variability, and library preparation biases. Current benchmarking approaches address these issues insufficiently: BUSCO assesses transcriptome completeness using conserved single-copy orthologs but can misinterpret alternative splicing as gene duplications, while spike-ins (SIRVs, ERCCs) oversimplify real- sample complexity, neglecting RNA degradation and RNA extraction artifacts, thus inflating performance metrics. Simulation algorithms are limited to recapitulate this complexity. To overcome these limitations, we introduce the Transcriptome Universal Single-isoform Control (TUSCO), a curated internal reference set of genes lacking alternative isoforms. TUSCO evaluates precision by identifying transcripts deviating from reference annotations and assesses sensitivity by verifying detection completeness in human and mouse samples. Masking TUSCO transcripts--and optionally inserting decoy splice variants--creates a novel- isoform challenge that assesses recovery of the true, now-unannotated isoforms. Our validation demonstrates that TUSCO provides accurate and reliable benchmarking without external controls, significantly improving quality control standards for transcriptome reconstruction using LRS.

bioinformatics↗

bioRxiv2025

CASP3-GSDME-mediated trophoblast pyroptosis contributes to systemic inflammation

Early-onset preeclampsia (EOPE) is associated with excessive apoptosis and inflammation, but the mechanistic link between these processes remains enigma. Here, we report elevated circulating pro-apoptotic proteins in EOPE patients at early pregnancy, along with concurrent CASP3 activation and GSDME cleavage in a subset of EOPE placentas. Using multiple trophoblast cell lines, we demonstrate that trophoblast cells, which highly express GSDME, undergo a shift from apoptosis to CASP3-dependent pyroptosis, driving inflammation. Notably, pyroptotic trophoblasts further induce pro-inflammatory macrophage polarization within placental villi organoids, establishing a feedback loop that amplifies both trophoblast pyroptosis and inflammatory responses in trophoblast organoids-macrophage assembloids. In vivo, CASP3-GSDME-mediated trophoblast pyroptosis contributes to systemic inflammation in wild-type pregnant mice but not in Gsdme-/- mice. Screening of EOPE prevention drugs reveals Vitamin D as a suppressor of GSDME activation and pyroptosis in trophoblast cells. These findings highlight the CASP3-GSDME axis as a promising therapeutic target for preeclampsia prevention.

cell biology↗

bioRxiv2025

Identification and quantification of alternative polyadenylation sites in single cell RNA-seq data using scPAISO

AbstractsAlternative polyadenylation (APA) is a critical posttranscriptional mechanism that generates transcriptomic diversity through the production of mRNA isoforms with distinct 3 UTRs or coding sequences. Current APA analysis based on single-cell RNA sequencing (scRNA-seq) for establishing cell type-specific APA landscapes primarily rely on Read2 data, which lacks precise cleavage site (CS) information. This limitation restricts their ability to achieve precise de novo mapping of polyadenylation sites (PASs). Here, we present single-cell PolyAdenylation ISOform quantification (scPAISO), a computational pipeline designed for de novo identification of PAS and quantification of PAS isoforms in scRNA-seq data, by leveraging the often-discarded Read1 from 3 tag-based scRNA-seq protocols. Unlike existing tools, scPAISO directly captures mRNA 3 end cleavage sites, enabling superior performance in motif enrichment (stronger AAUAAA signal) and peak precision (sharper PAS peaks). Moreover, the smaller peak widths enhance the spatial resolution, enabling more accurate detection of closely spaced PASs in the genome. By integrating Read1 and Read2 data, scPAISO achieves isoform-level quantification with an assignment accuracy exceeding 95%. We demonstrate the robustness of scPAISO in identifying PASs and quantifying APA events across diverse biological contexts, including hematopoiesis, systemic sclerosis (SSc), and mouse tissues. We identified stage-specific 3 UTR lengthening in hematopoietic progenitors, global 3 UTR remodeling in SSc and tissue-specific polyadenylation (PA) preference along with RNA-binding proteins in mice. Overall, scPAISO represents a significant advancement in the analysis of APA at single-cell resolution and provides a powerful tool for exploring the regulatory landscape of APA, offering new insights into transcriptome complexity and gene regulation in both health and disease.

bioinformatics↗

bioRxiv2025

Structural connectome architecture and biological vulnerability shape cortical atrophy in cocaine use disorder

Cocaine use disorder (CUD) is prevalent and characterized by widespread gray matter atrophy across the cerebral cortex. However, it remains unclear whether and how connectome-based circuits and biological features shape these structural abnormalities. Here, we show that CUD-associated regional atrophy is constrained by the white matter (WM) structural connectome. Along these WM pathways, regions that share similar haemodynamic activity and molecular profiles are more likely to exhibit analogous atrophy patterns. By integrating the structural connectome with these multiple connectivity blueprints, we subsequently identify CUD epicenters and reveal that the prefrontal and visual cortices serve as core systems. Furthermore, we link these epicenters to cortical gene expression patterns and delineate CUD-related transcriptomic enrichment in biological processes, including inter-neuronal communication, synaptic function and neural homeostasis. Finally, we demonstrate that the spatial distribution of these epicenters correlates with cocaine craving-response maps derived from repeated transcranial magnetic stimulation and can track individual variations in clinical behavioural representations, suggesting their potential as key targets for therapeutic intervention. Together, our findings establish a structurally constrained framework for the spread of pathology underlying cortical atrophy in CUD, where initial perturbations propagate via structural connectome pathways to vulnerable regions shaped by neural activity and molecular landscapes.

neuroscience↗

bioRxiv2025

The nucleotide analog bemnifosbuvir inhibits hepatitis E virus replication in preclinical models

BackgroundHepatitis E virus infections remain a global health concern. Immunocompromised patients are at an increased risk to develop chronic HEV infection and thereby severe liver disease. Current off-label regimens are suboptimal with treatment failure being reported. Therefore, there is an urgent need for an effective anti-HEV treatment. ObjectiveIn this study, we aimed to identify potent inhibitors of HEV replication. DesignWe developed a rapid, image-based screening platform based on a full-length HEV fluorescence reporter virus and screened a nucleotide/nucleoside analog library. The identified lead candidate was validated in authentic hepatocyte culture systems, as well as in a gerbil infection model. ResultsBemnifosbuvir (BEM), previously characterized as a nucleotide analog with activity against other RNA viruses, efficiently suppressed HEV replication in vitro and in vivo in a dose-dependent manner, with minimal cytotoxicity at effective concentrations. Combining BEM with ribavirin, the off-label drug given to chronic HEV patients, resulted in an additive antiviral effect against HEV. We found that HEV-3 remains susceptible to inhibition by BEM over an extended treatment period, reducing concerns about the rapid development of viral resistance. Importantly, BEM significantly reduced HEV viral loads and liver inflammation in a gerbil infection model. ConclusionsGiven BEMs favorable safety profile in preclinical and clinical settings, our results suggest investigating its efficacy in patients with chronic HEV infection.

microbiology↗

bioRxiv2025

Betrixaban Activates cGAS and ERVs to Promote Dual Nucleic-Sensing Antiviral Immunity

Broad-spectrum host-directed antivirals are urgently needed, as virus-targeted drugs often suffer from narrow specificity and rapid resistance. Here, we reported that Betrixaban (BT), an FDA-approved oral Factor Xa inhibitor, induced a robust antiviral state through dual innate immune pathways. Mechanistically, we identified BT directly bound and activated the DNA sensor cGAS to induce cGAMP production, representing BT as the first small-molecule cGAS agonist. Concurrently, BT inhibited histone deacetylases (HDACs), leading to chromatin de-repression of endogenous retroviruses (ERVs) and production of immunostimulatory double-stranded RNA (dsRNA) that engaged RIG-I/MDA5. These combined signal cascades triggered strong type I interferon responses and conferred broad-spectrum antiviral protection against RNA and DNA viruses in vitro and in vivo. These findings unveil a unique host-directed antiviral strategy wherein a small molecule drug engage dual nucleic acid-sensing pathway, and suggest repurposing BT as an orally broad-spectrum antiviral. Graphical abstract O_FIG O_LINKSMALLFIG WIDTH=200 HEIGHT=190 SRC="FIGDIR/small/669242v1_ufig1.gif" ALT="Figure 1"> View larger version (68K): org.highwire.dtl.DTLVardef@aa153dorg.highwire.dtl.DTLVardef@585d43org.highwire.dtl.DTLVardef@152ec59org.highwire.dtl.DTLVardef@11ada33_HPS_FORMAT_FIGEXP M_FIG C_FIG

immunology↗

bioRxiv2025

Improving polygenic risk prediction performance through integrating electronic health records by phenotype embedding

Large-scale biobanks provide comprehensive electronic health records (EHRs) that capture detailed clinical phenotypes, potentially enhancing disease risk prediction. However, traditional polygenic risk score (PRS) methods rely on simplified phenotype definitions or predefined trait sets, limiting their ability to represent the intricate phenotypic structures embedded within EHRs. To address this gap, we introduce a general framework, EEPRS, that leverages phenotype embeddings derived from EHRs to improve genetic risk prediction using only genome-wide association study (GWAS) summary statistics, enabling accurate, robust and interpretable risk prediction for a wide range of diseases. Employing embedding methods such as Word2Vec and GPT, we conducted EHR embedding-based GWAS and identified a distinct cardiovascular cluster via hierarchical clustering of genetic correlations. Across 41 clinical traits in the UK Biobank, our EEPRS framework consistently outperformed traditional single-trait PRS, particularly within this identified cluster. Validation using PRS-based phenome-wide association studies (PRS-PheWAS) further confirmed robust associations between EHR embedding-based PRS and circulatory system diseases. Furthermore, our data-adaptive method, EEPRS_optimal, employing cross-validation to select the best embedding method, leading to additional improvements in prediction. We further developed MTAG_EEPRS for multi-trait PRS, resulting in averaging 92.48% improvement in R2 for continuous traits and 24.06% in AUC for binary traits compared to single-trait PRS. Overall, EEPRS represents a robust and interpretable framework, enhancing genetic prediction accuracy through integrating EHR embeddings with single-trait and multi-trait PRS.

genetics↗

bioRxiv2025

Target-Stabilized Base Editors Enable Robust High-Fidelity RNA Editing

RNA base editing using engineered deaminases expands target compatibility, offering a powerful tool to correct mutations at the RNA level. However, widespread off-target effects, primarily arising from dissociated free deaminases, remain a significant challenge. Here, we devised the RECODE (RNA editing with conditionally stable and enhanced ADAR1 deaminase variants) system, which employs designer degron-tagged ADAR1 deaminase (ADAR1d) with guide RNA (gRNA)-regulated stability. By promoting degradation of gRNA-unbound ADAR1d, RECODE markedly reduced transcriptome-wide edits while maintaining high on-target efficacy. Engineering gRNA for target RNA-induced conformational switching restricts ADAR1d stabilization, further enhancing editing precision. With structure-guided rational engineering of ADAR1d, RECODE achieved editing efficiencies of 90.3% and 58.8% for an Amyotrophic Lateral Sclerosis-relevant FUS mutation in vitro and in vivo, respectively. These findings establish RECODE as a highly stringent and efficient RNA editing technology and underscore a general principle for enhancing the specificity of RNA-guided protein effectors.

molecular biology↗

bioRxiv2025

Human whole epigenome modelling for clinical applications with Pleiades

Gene regulation in humans extends beyond the four-letter genetic code. DNA methylation, in particular, functions as a critical epigenetic regulator, dynamically programming cellular identity, adapting gene expression in response to environmental cues, and underpinning the onset and progression of numerous diseases. Here we present Pleiades, a series of whole-genome epigenetic foundation models spanning three sizes: 90M, 600M, and 7B parameters. Pleiades is trained upon an extensive proprietary corpus of human methylation and genomic data, totalling 1.9T tokens. We introduce alignment embeddings and stacked hierarchical attention techniques to provide precise epigenetic modelling without the need for extended context lengths. Collectively, these advances enable Pleiades to perform a diverse range of down-stream biological and clinical tasks, including genomic regulatory prediction, realistic generation of cell-free DNA fragments and fragment-level cell-type-of-origin classification, within a unified and scalable computational framework. We specifically apply Pleiades to the early detection of clinical Alzheimers disease and Parkinsons disease from plasma cell-free DNA, achieving high-accuracy detection (AUROC 0.89 for AD and 0.84 for PD) using a minimally invasive blood test. Combined with plasma pTau-217, Pleiades reaches an AUROC of 0.97 for AD, underscoring the promise of multimodal epigenomic and proteomic approaches. Using mechanistic interpretability, we ground Pleiades latent features in interpretable biological signals, relating its Alzheimers predictions to fragmentomic and epigenomic properties of cfDNA. These findings demonstrate the potential of genome-wide epigenomic modelling as a clinically translatable paradigm for diagnostics and precision medicine, though prospective validation in larger, more diverse cohorts will be required to confirm clinical utility.

genomics↗

bioRxiv2025

Cell type-specific master metabolic regulators of Alzheimer's disease

Alzheimers disease (AD) exhibits metabolic heterogeneity; yet, the consequences on metabolic dynamics in a cell-type-specific manner and the underlying metabolite-sensor network basis remain unclear. Here, we show that neurons exhibit a striking decrease in energy and lipid-related metabolic activity, contrasted by an increase in microglial metabolism associated with neuroinflammation. To identify cell-type specific master metabolic regulators of AD underlying the metabolic alterations in AD, we introduce scFUMES (single cell FUnctional MEtabolite-Sensor), an algorithm integrating single-cell RNA sequencing, interactomics, genomics, transcriptomics, and metabolomics from human brain biobanks. Applied to two AD-vulnerable regions (middle temporal gyrus and dorsolateral prefrontal cortex), scFUMES uncovers hundreds of AD-associated regulators, with neurons and microglia showing the most interactions. Particularly, scFUMES pinpoints genetics-informed master metabolic regulators across AD severity, sex and APOE genotype (e.g., PPARD-glycerol in microglia). Experimental testing reveals that two interaction pairs predicted by scFUMES, neuronal palmitic acid bound fatty acid binding protein 3 and gut metabolite indole-3-propionic acid binding to kynurenine aminotransferase 1, both lower pathological tau species in AD. Collectively, scFUMES systematically maps AD master metabolic regulators, offering insights into cellular metabolic heterogeneity and therapeutic strategies for AD and other AD-related dementia if broadly applied.

systems biology↗

bioRxiv2025

Statistically valid explainable black-box machine learning: applications in sex classification across species using brain imaging

Sex classification using neuroimaging data has the potential to revolutionize personalized diagnostics by revealing subtle structural brain differences that underlie sex-specific disease risks. Despite the promise of machine learning, traditional methods often fall short in providing both high classification accuracy and interpretable, statistically validated feature importance scores for high-dimensional imaging data. This gap is particularly evident when conventional techniques such as random forests, LIME, and SHAP are applied, as they struggle with complex feature interactions and managing noise in large datasets. We address this challenge by developing an integrated framework that combines Oblique Random Forests (ORFs) with a novel, permutation-based feature importance testing algorithm. ORFs extend traditional random forests by employing oblique decision boundaries through linear combinations of features, thereby capturing intricate interactions inherent in neuroimaging data. Our feature importance testing method, NEOFIT, rigorously quantifies the significance of each feature by generating null distributions and corrected p-values. We first validate our approach using simulated datasets, establishing its robustness and scalability under controlled conditions. We then apply our method to classify sex from both voxel-wise structural MRI and cortical thickness data in humans and macaques, facilitating direct cross-species comparisons. ORFs achieves AUC > 0.80 on human data, and > 0.70 on macaque data, while NEOFIT identifies statistically significant features aligned with sex-dimorphic neuroanatomy. Our results demonstrate that the proposed framework not only enhances classification performance but also provides clear, interpretable insights into the neuroanatomical features that distinguish sexes. These methodological advancements pave the way for improved diagnostic tools and contribute to a deeper understanding of the evolutionary basis of sex differences in brain structure.

neuroscience↗