Biology subjects

Zhi, D.

Publications and source records attributed to Zhi, D..

At least 19 recordsLinked to original sources

bioRxiv2026

Inflections in adolescent cortical maturation signal psychopathology risk

Cerebral cortical maturation supports emotional, social, and cognitive development during adolescence1. Neurodevelopmental theories of psychiatric illnesses, which peak in onset around age 182, implicate subtle alterations in adolescent cortical development3. However, mechanistic links between cortical maturation and emergent psychopathology have remained elusive. This gap likely reflects a need for more nuanced understanding of cortical development4, larger, longitudinal samples powered to detect small effects,5 and more uniform image quality control.6 Using 29,482 structural MRI scans from 11,720 adolescents in the Adolescent Brain Cognitive Development (ABCD) study, we demonstrate reproducible differences in the contour and tempo of maturation across the cortical surface. Regional trajectories are robust to image quality variance and align with increasing functional specialization over time. Thicker cortex in early-maturing sensory regions predicts psychiatric symptoms as early as four years prior to their emergence, a pattern that subsequently spreads to association cortex. Overlap with spatially and temporally organized gene co-expression networks implicates prenatal programming of this sequence of maturation and its links to psychopathology. These findings present a novel, versatile neurodevelopmental framework for cortical maturation during adolescence that sequentially implicates sensory and association cortex in risk for emergent psychiatric symptoms.

neuroscience↗

bioRxiv2026

Recomb-Mix Cloud: a cost-efficient and scalable cloud-friendly pipeline for local ancestry inference

Local ancestry inference (LAI) reveals population compositions of an individuals chromo-somes, which empowers population genetics and genetic association studies of diverse populations. The recent availability of large genomic datasets from biobanks offers unprecedented opportunities for LAI. However, such datasets are typically only accessible from cloud-based platforms, and the lack of efficient and cost-effective tooling makes LAI prohibitive for resource-constrained researchers. Here, we develop Recomb-Mix Cloud, a cloud-ready pipeline for LAI based on Recomb-Mix, a recently introduced efficient LAI method. Recomb-Mix focuses on individual variants while maintaining a lower computational burden than other LAI methods focusing on genomic windows. Recomb-Mix Cloud offers a cloud-friendly complete pipeline from panel sample selection, marker alignment, compact reference panel construction, Recomb-Mix threading, and generating LAI calls in common data formats. Recomb-Mix Cloud also offers a set of pre-computed compact population reference panels constructed based on data from the Human Genome Diversity Project and the 1000 Genomes Project. Our results showed Recomb-Mix is about one and two orders of magnitude cheaper and faster than FLARE and RFMix, two popular LAI methods, while offering largely consistent LAI calls on the All of Us Researcher Workbench. The estimated overall compute cost per All of Us sample for whole-genome LAI analysis using Recomb-Mix Cloud with a pre-computed reference panel is just $0.01, compared with $0.43 for FLARE and $4.58 for RFMix. Therefore, Recomb-Mix Cloud is well-positioned to democratize local ancestry-based genetics research on modern biobank-sized data. The Recomb-Mix Cloud pipeline and the compact reference panels are available at https://github.com/ucfcbb/Recomb-Mix.

bioinformatics↗

bioRxiv2026

Sum-h2, enabling genetic discovery for deep learning-derived phenotypes through a fast evaluation framework and arena of performance

In the recent growing interest of AI research toward biology, genetic association studies of AI- derived phenotypes from high-content modalities such as images emerges as a powerful means for biological discovery. However, such AI-phenotyping methods still lacks a good optimization target and an efficient evaluation framework. The number of discovered loci was the major criterion for evaluating the quality of AI-derived phenotypes. However, the time and computational resources required for running the GWAS and subsequent loci-clumping are substantial, limiting the rapid development and iteration of deep learning representation algorithms. Here we present a 1000x faster and lightweight framework, sum-h2, than traditional GWAS framework for evaluating genetic discovery through total heritability. We revisit tr(P^(-1) G), previously proposed in the context of evolutionary studies, as a measure of multi-phenotype heritability. We showed that sum-h2, the sum of heritability over phenotypic Principal Components (PCs), is equivalent to the linear transformation-invariant tr(P^(-1) G), through both theoretical proof and simulation studies. Moreover, sum-h2 can be estimated rapidly with minimal information loss over a relatedness-enriched sample, while preserving the relative ranking of endophenotypes by GWAS loci counts. Based on selected UKB data and sum-h2, we set up a Genetic Discovery Arena, enabling rapid and fair comparisons for the development and optimization of deep learning-derived phenotyping methods.

bioinformatics↗

bioRxiv2026

PerturbLDM: conditional latent diffusion for modelling single-cell perturbation responses

Single-cell perturbation profiling maps intervention-induced phenotypes, yet experiments measure only a fraction of the perturbation-context space. Learning context-dependent perturbation effects could enable response prediction beyond measured conditions. Here we introduce PerturbLDM, a latent-diffusion framework for conditional generation of single-cell transcriptional responses. Following Tahoe-100M pretraining, it predicted 13,942 held-out combinations of observed drugs, doses and cell lines more accurately than existing methods, with higher matched-control effect correlation than an additive marginal baseline in 95.2% of conditions. The Tahoe-100M-pretrained model was further used to rank PANACEA compounds by pathway similarity, placing shared-mechanism pairs among nearest neighbours. In smaller datasets, Per-turbLDM generated a mid-gestational fetal-colon state with 67% lower gene-wise error than Squidiff, retaining the balance between absorptive and BEST4/OTOP2-like epithelial programmes. In PBMCs, it captured six of seven interferon and antiviral programmes and the interferon-associated FAO-OXPHOS programme more accurately than scGen. Together, these results support conditional response generation across data scales and biological settings.

bioinformatics↗

bioRxiv2026

Individualized parcellation reveals functional boundaries in human prefrontal cortex

Human prefrontal cortex (PFC) supports diverse higher-order cognitive functions during task execution. Increasing evidence suggests that these functions are organized along large-scale gradients, such as a rostro-caudal axis supporting progressively abstract-to-concrete information processing. At the same time, regional specialization within PFC, including focal patches selective for specific stimulus categories, suggests the presence of discrete functional boundaries. Whether PFC organization is best characterized by continuous gradients or discrete subdivisions remains unresolved. Here we show that task-evoked functional organization in PFC exhibits substantial inter-individual variability, limiting the usefulness of conventional group atlases in testing for functional boundaries. To address this challenge, we estimated individualized functional parcellations by combining a group atlas with task-evoked fMRI data spanning diverse cognitive tasks. The group atlas revealed large-scale functional gradients across PFC, whereas individualized parcellations additionally uncovered sharp functional boundaries obscured by group averaging. Moreover, functional organization in PFC was substantially more fine-grained than in other association cortices, consistent with its integrative role in cognitive control. Together, these findings suggest that PFC organization reflects an individualized mosaic of fine-grained functional subdivisions embedded within broader large-scale gradients, with important implications for defining core constructs such as the multiple-demand system.

neuroscience↗

bioRxiv2026

Mapping the genetic architecture of human cortical expansion and its links to neuropsychiatric disorders

Cortical expansion is a defining feature of human evolution and neurodevelopment, involving the tangential growth and gyrification of the cerebral cortex. Although disruptions to the expansion of the cortex are implicated in diverse brain-based disorders, the genetic architecture underlying this process remains undercharacterized. Here, we integrated GWAS data for five neuroanatomical phenotypes measured in up to 73,800 individuals to model a latent genomic factor capturing the shared genetic basis of cortical surface area, folding, curvature, gyrification, and intracranial volume. Using a multivariate framework, we mapped this pleiotropic architecture across biological scales, identifying novel effector genes, neural cell types, and developmental pathways involved in cortical expansion. Functional genomic evidence indicated that a diverse cellular ensemble contributes to the tangential growth and gyrification of the cortex during prenatal neurodevelopment, with progenitor cell lineages prominently involved. Regional analyses further revealed that genetic influences are spatially heterogeneous and organized along canonical anatomical, functional, and developmental gradients. Finally, cortical expansion exhibited substantial genetic overlap with neurodevelopmental, psychiatric, and neurological disorders, with shared genetic influences often concentrated in prefrontal cortical regions. Together, these findings map the genetic architecture of human cortical expansion and implicate it as a common etiological axis linking evolution, neurodevelopment, and health and disease.

genomics↗

bioRxiv2026

Precision Functional Parcellation of the Human Cortex via Rest-Task fMRI Fusion

Individual-specific cortical parcellations enable the characterization of brain network organization that is often obscured by population-level atlases, with broad implications for both basic neuroscience and translational applications. However, existing methods rely primarily on resting-state fMRI and underutilize task-evoked data, which provide complementary information about functional specialization. This limitation partly reflects the challenge of integrating heterogeneous datasets that differ in task design, sample size, and cortical coverage. Here, we present mRBM-HBP, a scalable hierarchical Bayesian framework that incorporates a multinomial restricted Boltzmann machine to model spatial dependencies, enabling efficient and flexible integration of resting-state and task fMRI across diverse datasets and inference of both group-level and individual-level cortical parcellations. We show that mRBM-HBP achieves performance comparable to state-of-the-art resting-state-based parcellation methods while substantially reducing computational cost. By integrating large-scale task-fMRI datasets, we derive a task-based parcellation and demonstrate that resting-state and task conditions reveal largely consistent macroscopic networks, while task data provide state-specific refinements of functional boundaries. Moreover, a fused rest-task group-level atlas improves the accuracy, reliability, and individual specificity of inferred parcellations, particularly when individual-level data are limited. These results indicate that integrating resting-state and task fMRI enhances precision mapping of functional brain organization.

neuroscience↗

bioRxiv2026

Replicability of unsupervised deep learning derived image phenotypes

Unsupervised deep-learning image phenotypes derived from brain MRI are propelling imaging genetics to link brain structure to genetic variation. However, their replicability across data sets has not been sufficiently evaluated, raising questions about whether they capture robust biological structure or reflect training-specific artifacts. Here, we assess the replicability of unsupervised deep-learning image phenotypes under variation in model initialization, data partitioning, and cohort, directly evaluating their stability across experimental conditions. We trained multiple models under (i) different training batch random seeds, (ii) cross-validation splits, and (iii) independent datasets (UKB and ADNI), across CNN and ViT architectures. We then derived representations from a separate UKB discovery cohort (N = 22,985) for both trained models and random initialized models without training. The representation stability was assessed using centered kernel alignment (CKA; mean ViT 0.74 vs random 0.27) and kernel canonical correlation analysis (KCCA; mean ViT 0.84 vs random 0.60), as well as genetic discovery stability using loci overlap ratio (mean ViT 0.45 vs random 0.08). We further applied weighted MAXVAR generalized CCA to 12 embeddings to extract a shared 30-dimensional subspace. Our result showed that UDIPs exhibit statistically significant stability (CKA, KCCA t test p < 0.001) across training perturbations and preserve biologically meaningful structure (loci overlap ratio t test p <0.001) across cohorts, supporting their use in imaging genetics.

bioinformatics↗

bioRxiv2026

Learning dynamics of unsupervised deep learning reveal epoch-specific genetic architectures of brain morphology

Representation learning is an emerging paradigm for deriving phenotypes from complex measurements (e.g., imaging) for genetic discovery. However, the learning dynamics of deep neural networks, especially the evolution of representations during training, while of interest in representation learning, were insufficiently investigated in the context of genetic discovery. In this study, using a 3D convolutional autoencoder trained on T1-weighted brain MRIs UK Biobank participants, we show that its learning trajectory forms an epoch-stratified landscape of brain morphology heritability. Different training epochs capture distinct genetic architectures at comparable heritability levels. Overall, ensembling across informative checkpoints identifies more genomic risk loci than the conventional single-checkpoint approach. Interpretability analysis reveals that epoch-specific loci, including MAPT and MCPH1, map onto biologically coherent and distinct neuroanatomical signatures, identified at different stages of the training process. Our results establish learning dynamics as a novel axis for genetic discovery using unsupervised deep learning and have practical implications for any architecture that saves multiple checkpoints during training.

bioinformatics↗

bioRxiv2026

SLAB: A Sweep Line Algorithm in PBWT for Finding Haplotype Block Cores

With the increasing availability of high-quality phased haplotype data, researchers can more effectively identify detailed patterns of haplotype sharing and investigate the population genetic processes that shape them. In this work, we define block cores as genomic segments where multiple haplotype blocks overlap. We develop an efficient algorithm to analyze haplotype blocks, focusing on width-maximal matches and the identification of haplotype block cores. We apply our algorithm to UK Biobank haplotypes to quantify block cores and demonstrate the biological and population-level insights that can be inferred from their patterns. Specifically, identified block cores can serve as a basis for detecting signals of selection. Although our results are largely consistent with those from methods such as IBD rate analysis, they also reveal complementary information not captured by IBD rates or other conventional approaches. Source code is available at https://github.com/ZhiGroup/SLAB.

bioinformatics↗

bioRxiv2026

Learning heritable multimodal brain representation via contrastive learning

Magnetic resonance imaging (MRI)-derived phenotypes (IDP) has enabled the discovery of numerous genomic loci associated with brain structure and function. However, most existing IDPs and learned representations are derived from a single imaging modality, missing complementary information across modalities and potentially limiting the scope of genetic discovery. Here, we introduce a multimodal contrastive learning framework to derive heritable representations from paired T1- and T2-weighted MRIs. Unlike single-modality reconstruction-based models, we designed a momentum-based contrastive learning framework. As a result, our approach offers improved prediction of traditional IDPs, age, and brain disorders. Notably, genome-wide association studies (GWAS) of the learned representations reveal a substantially higher overlap of genetic loci across modalities, indicating improved alignment of their underlying genetic architecture. Analysis of the GWAS loci identified shared protein and drug targets, yielding meaningful biological insights. Overall, our framework learns shared representations across brain imaging modalities that exhibit anatomical and genetic coherence.

bioinformatics↗

bioRxiv2026

RLBWT-Based LCP Computation in Compressed Space for Terabase-Scale Pangenome Analysis

Lossless full text indexes are utilized in a myriad of applications in bioinformatics. The continuously decreasing cost of generating biological data has resulted in the need to build full text indexes on biological datasets of increasing size. Many compressed full text indexes have been developed to address this problem. In particular, run-length Burrows-Wheeler transform (RLBWT) based compressed full text indexes have seen wide development and adoption. However, the construction of these RLBWT-based compressed full text indexes is still computationally expensive, sometimes prohibitively so, even for current dataset sizes. Therefore, we present algorithms for the construction of RLBWT-based compressed full text indexes and their supporting data structures in compressed space. The algorithms have a space complexity of O(r) words and run in O(n) time for repetitive datasets, where r is the number of runs in the BWT, n is the length of the text, and repetitive datasets implies the average run length is at least log n. We provide the first algorithm to compute LCP-related information for repetitive datasets in optimal time and O(r) space, greatly reducing memory requirements. The key idea behind this algorithm is the utilization of r samples of the inverse suffix array at regular intervals. For example, on the Human Pangenome Reference Consortium Release 2 dataset, this reduces peak memory from 2,135 GiB to 170 GiB (12.6x reduction) compared to the previous best method (pfp-thresholds). AvailabilityThe implementation is available at https://github.com/ucfcbb/TeraTools. Supplementary InformationSupplementary Material is available online at bioRxiv.

bioinformatics↗

bioRxiv2025

Crosstalk between Immune Microenvironment and Hair Follicle Cells Underlies Sexual Dimorphism in Androgenetic Alopecia

Androgenetic alopecia (AGA), also known as female pattern hair loss (FPHL) in women, is the most prevalent form of hair loss. It is characterized by progressive miniaturization of hair follicles and shortening of the anagen phase. The condition is widely attributed to genetic predisposition and androgen-mediated activation of androgen receptors. Other factors, such as the immune microenvironment, could also contribute to the pathogenesis. However, the specific mechanisms involved are still poorly understood. This study aimed to investigate the potential role of skin-resident myeloid cells in interacting with hair follicle cells under androgen stimulation, and to elucidate the sex-specific differences in dihydrotestosterone (DHT)-induced hair loss. Both female and male mice received DHT treatment, and histological examination was performed to assess DHT-induced alterations in skin morphology. Single-cell RNA sequencing (scRNA-seq) was utilized to profile skin cell populations and explore underlying mechanisms. Our results demonstrated that DHT inhibited hair regrowth in both sexes, altered skin thickness, and induced hair follicle miniaturization. scRNA-seq analysis revealed enhanced interactions between myeloid and fibroblast subpopulations, with more robust crosstalk observed in female mice. An in vitro experiment demonstrated that DHT promoted apoptosis of dermal papilla cells (DPCs) in the presence of macrophages. Treatment with polypeptides Y001 and Y002 effectively promoted hair regrowth by suppressing apoptosis signaling pathways. Our findings underscore the interactions between immune cells and hair follicular cells, particularly mediated by skin-resident myeloid cells, in the sexual dimorphism of androgenetic alopecia. The polypeptides Y001 and Y002 exhibit promising therapeutic potential by targeting apoptotic pathways, offering novel avenues for AGA treatment.

cell biology↗

bioRxiv2025

Uncovering genetic architecture of the heart via genetic association studies of unsupervised deep learning derived endophenotypes

Recent genome-wide association studies (GWAS) have effectively linked genetic variants to quantitative traits derived from time-series cardiac magnetic resonance imaging, revealing insights into cardiac morphology and function. Deep learning approach generally requires extensive supervised training on manually annotated data. In this study, we developed a novel framework using a 3D U-architecture autoencoder (cineMAE) to learn deep image phenotypes from cardiac magnetic resonance (CMR) imaging for genetic discovery, focusing on long-axis two-chamber and four-chamber views. We trained a masked autoencoder to develop Unsupervised Derived Image Phenotypes for heart (Heart-UDIPs). These representations were found to be informative to indicate various heart-specific phenotypes (e.g., left ventricular hypertrophy) and diseases (e.g., hypertrophic cardiomyopathy). GWAS on Heart UDIP identified 323 lead SNP and 628 SNP-prioritized genes, which exceeded previous methods. The genes identified by method described herein, exhibited significant associations with cardiac function and showed substantial enrichment in pathways related to cardiac disorders. These results underscore the utility of our Heart-UDIP approach in enhancing the discovery potential for genetic associations, without the need for clinically defined phenotypes or manual annotations.

bioinformatics↗

bioRxiv2025

Deep Learning to Predict Future Cognitive Decline: A Multimodal Approach Using Brain MRI and Clinical Data

Predicting the trajectory of clinical decline in aging individuals is a pressing challenge, especially for people with mild cognitive impairment, Alzheimers disease, Parkinsons disease, or vascular dementia. Accurate predictions can guide treatment decisions, identify risk factors, and optimize clinical trials. In this study, we compared two deep learning approaches for forecasting changes, over a 2-year interval, in the Clinical Dementia Rating scale sum of boxes score (sobCDR). This is a key metric in dementia research, and scores range from 0 (no impairment) to 18 (severe impairment). To predict decline, we trained a hybrid convolutional neural network that integrates 3D T1-weighted brain MRI scans with tabular clinical and demographic features (including age, sex, body mass index (BMI), and baseline sobCDR). We benchmarked its performance against AutoGluon, an automated multimodal machine learning framework that selects an appropriate neural network architecture. Our results demonstrate the importance of combining image and tabular data in predictive modeling for clinical applications. Deep learning algorithms can fuse image-based brain signatures and tabular clinical data, with potential for personalized prognostics in aging and dementia.

neuroscience↗

bioRxiv2025

Haplotype-based Parallel PBWT for Biobank Scale Data

Durbins positional Burrows-Wheeler transform (PBWT) enables algorithms with the optimal time complexity of O(MN) for reporting all vs all haplotype matches in a population panel with M haplotypes and N variant sites. However, even this efficiency may still be too slow when the number of haplotypes reaches millions. To further reduce the run time, in this paper, a parallel version of the PBWT algorithms is introduced for all versus all haplotype matching, which is called HP-PBWT (haplotype-based parallel PBWT). HP-PBWT parallelly executes the PBWT by splitting a haplotype panel into blocks of haplotypes. HP-PBWT algorithms achieve parallelization for PBWT construction, reporting all versus all L-long matches, and reporting all versus all set-maximal matches while maintaining memory efficiency. HP-PBWT has an [Formula] time complexity in PBWT construction, and an [Formula] time complexity for reporting all versus all L-long matches and reporting all versus all set-maximal matches, where T is the number of threads and c* is the maximum number of matches (of length L or maximum divergence value for L-long matches and set-maximal matches, re-spectively) per haplotype per site. HP-PBWT achieves 4-fold speed-up in UK Biobank genotyping array data with 30 threads in the IO-included benchmarks. When applying HP-PBWT to a dataset of 8 million randomized haplotypes (random binary strings of equal length) in the IO-excluded benchmarks, it can achieve a 22-fold speed-up with 60 cores on the Amazon EC2 server. With further hardware optimization, HP-PBWT is expected to handle billions of haplotypes efficiently.

bioinformatics↗

bioRxiv2025

Dynamic μ-PBWT: Dynamic Run-length Compressed PBWT for Biobank Scale Data

Durbins positional Burrows-Wheeler transform (PBWT) supports efficient haplotype matching and queries given a panel of haplotypes. It has been widely used for statistical phasing, imputation and identity-by-descent (IBD) detection. However, the original PBWT panel doesnt support dynamic updates when haplotypes need to be added or deleted from the panel. Dynamic-PBWT (d-PBWT) solved this problem but it is not memory efficient. While the memory constraint problem of the PBWT has been tackled by Syllable-PBWT and {micro}-PBWT, these are static data structures that do not allow updates. Additionally, Syllable-PBWT only supports long-match query and {micro}-PBWT only supports set-maximal match query, limiting their functionality in the compressed form. In this paper, we present Dynamic {micro}-PBWT (which can also be seen as compressed d-PBWT) that is memory efficient and supports dynamic updates. We run-length compress PBWT to achieve better compression rate and store the runs in the self-balancing trees to enable dynamic updates. We show that the number of updates per insertion or deletion in the tree at each site is constant regardless of the number of haplotypes in the panel and the updates can be made without decompressing the index. In addition, we use orders of magnitude less memory than d-PBWT. We also provide a long match query algorithm that can easily be extended back to the original {micro}-PBWT. Overall, the flexibility and space-efficiency of Dynamic {micro}-PBWT makes it a potential index data structure for biobank scale genetic data analyses. The source code for Dynamic {micro}-PBWT is available at https://github.com/ucfcbb/Dynamic-mu-PBWT.

bioinformatics↗

bioRxiv2025

Haplotype Matching with GBWT for Pangenome Graphs

Traditionally, variations from a linear reference genome were used to represent large sets of haplotypes compactly. In the linear reference genome based paradigm, the positional Burrows-Wheeler transform (PBWT) has traditionally been used to perform efficient haplotype matching. Pangenome graphs have recently been proposed as an alternative to linear reference genomes for representing the full spectrum of variations in the human genome. However, haplotype matches in pangenome graph based haplotype sets are not trivially generalizable from haplotype matches in the linear reference genome based haplotype sets. Work has been done to represent large sets of haplotypes as paths through a pangenome graph. The graph Burrows-Wheeler transform (GBWT) is one such work. The GBWT essentially stores the haplotype paths in a run length compressed BWT with compressed local alphabets. Although efficient in practice count and locate queries on the GBWT were provided by the original authors, the efficient haplotype matching capabilities of the PBWT have never been shown on the GBWT. In this paper, we formally define the notion of haplotype matches in pangenome graph-based haplotype sets by generalizing from haplotype matches in linear reference genome-based haplotype sets. We also describe the relationship between set maximal matches, long matches, locally maximal matches, and text maximal matches on the GBWT, PBWT, and the BWT. We provide algorithms for outputting some of these matches by applying the data structures of the r-index (introduced by Gagie et al.) to the GBWT. We show that these structures enable set maximal match and long match queries on the GBWT in almost linear time and in space close to linear in the number of runs in the GBWT. We also provide multiple versions of the query algorithms for different combinations of the available data structures. The long match query algorithms presented here even run on the BWT in the same time complexity as the GBWT due to their similarity.

bioinformatics↗