Biology subjects

Wan, K.

Publications and source records attributed to Wan, K..

8 recordsLinked to original sources

bioRxiv2026

Information regimes in population-graph neuroimaging: cohort context and implementation-dependent evaluation sensitivity

Population-graph models aggregate a query subject with an unlabeled cohort, so their reported performance can depend on the evaluation regime rather than on predictive information alone. Evaluation context has two separable axes: the visibility of the cohort during training, and the context available at inference. We study the training-time axis under a protocol fixed before analysis, on an ABIDE-I cohort of 871 subjects (20 acquisition sites). Two population-graph implementations are compared: a learned/adaptive graph (EV-GCN) and a static weighted graph (Parisot-GCN). Within each implementation the cohort, connectivity features, outer splits and training and selection rules are held fixed, and only the information available to the graph varies. The primary estimand is the implementation-by-training-context interaction: the difference between the P0-to-P1 AUC changes of the two implementations ({Delta}{Delta}AUC). Across three preselected outer splits (666, 1024, 2024) the direction is consistent and positive, with observed interaction range 0.06285-0.13090; uncertainty is assessed separately within each split by paired subject and site-cluster bootstrap resampling and is reported per split. Splits are replication units: they are not pooled and no cross-split average is reported, so the result is not a claim of invariance to arbitrary splits. These findings tie reported population-graph performance to the evaluation information regime and motivate explicit reporting of cohort visibility. The contribution is methodological: a two-axis taxonomy of information regimes, preselected cross-split evidence that the sensitivity of reported performance to cohort context depends on the implementation, and a reusable reporting checklist.

neuroscience↗

bioRxiv2026

Tracing high transductive cohort AUC to same-site supervision in a site-aware population GNN for multisite fMRI

Population-graph models can use cohort-level context, complicating interpretation of strong multisite neuroimaging performance. We investigated which information pathways accounted for high transductive cohort discrimination in a site-aware heterogeneous population graph neural network for autism classification. Using a frozen ABIDE-I cohort (871 participants, 20 sites) and evaluation protocol, we applied controlled graph, feature, supervision, and architecture interventions to a clean-room re-implementation across C-PAC and NIAK preprocessing pipelines. Cohort out-of-fold AUC was approximately 0.94. A site-only graph retained similarly high discrimination (0.948 versus 0.941 for the full model), with no statistically significant difference detected. Removing same-site supervision under a site-wise intervention reduced AUC to 0.480, whereas a supervision-budget control retaining same-site supervision achieved 0.921. Among the alternative heads evaluated, the high cohort AUC was observed only with the sex-heterogeneous dual-channel head; a canonical topology-only Parisot-GCN did not reproduce it. A demographic classifier using site, sex, and their interaction achieved a best pooled AUC of 0.51. Under leave-one-site-out evaluation, AUC was 0.522 for C-PAC and 0.532 for NIAK, with confidence intervals including 0.5 and no clear evidence of unseen-site discrimination; an imaging-only reference achieved 0.653 and 0.588. Because leave-one-site-out evaluation jointly changes supervision availability, target-site topology, and training-domain composition, the decrease cannot be attributed to a single factor. These findings distinguish cohort-visible transductive performance from unseen-site generalization and motivate explicit site-only, no-graph, and site-held-out controls.

neuroscience↗

bioRxiv2023

A deep learning framework for predicting the neutralizing activity of COVID-19 therapeutics and vaccines against evolving SARS-CoV-2 variants

Understanding how viral variants evade neutralization is crucial for improving antibody-based treatments, especially with rapidly evolving viruses like SARS-CoV-2. Yet, conventional assays are limited in the face of rapid viral evolution, relying on a narrow set of viral isolates, and falling short in capturing the full spectrum of variants. To address this, we have developed a deep learning approach to predict changes in neutralizing antibody activity of COVID-19 therapeutics and vaccines against emerging viral variants. First, we trained a variational autoencoder (VAE) using all 67,885 unique SARS-CoV-2 spike protein sequences from the NCBI virus (up to October 31, 2022) database to encode spike protein variants into a latent space. Using this VAE and a curated dataset of 7,069 in vitro assay data points from the NCATS OpenData Portal, we trained a neural network regression model to predict fold changes in neutralizing activity of 40 COVID-19 therapeutics and vaccines against spike protein sequence variants, relative to their neutralizing activity against the ancestral strain (Wuhan-Hu-1). Our model also employs Bayesian inference to quantify prediction uncertainty, providing more nuanced and informative estimates. To validate the models predictive capacity, we assessed its performance on a test set of in vitro assay data collected up to eight months after the data included in the model training (N = 980). The model accurately predicted fold changes in neutralizing activity for this prospective dataset, with an R2 of 0.77. Expanding our methodology to include all available data from NCBI virus and NCATS OpenData Portal up to date, we assessed predicted changes in activity for current COVID-19 monoclonal antibodies and vaccines against newly identified SARS-CoV-2 lineages. Our predictions suggest that current therapeutic and vaccine-induced antibodies will have significantly reduced activity against newer XBB descendants, notably EG.5, FL.1.5.1, and XBB.1.16. Using the model, we were able to primarily attribute the observed predicted loss in activity to the F456L spike mutation found in EG.5 and FL.1.5.1 sequences. Conversely, mRNA-bivalent vaccines are predicted to be less susceptible to the recent BA.2.86 variant compared to new XBB descendants. These findings align closely with recent research, underscoring the potential of deep learning in shaping therapeutic and vaccine strategies for emerging viral variants.

immunology↗

bioRxiv2023

Integration of eQTL and Machine Learning Methods to Dissect Causal Genes with Pleiotropic effects in Genetic Regulation Networks of Seed Cotton Yield

Expression quantitative trait loci (eQTL) provide a powerful means of investigating the biological basis of genome-wide association study (GWAS) results and exploring complex traits or phenotypes. In addition to identifying the causal gene in cis, eQTL analysis also reveals a large number of trans-regulated genes located on different chromosomes, which form a gene regulatory network (GRN) that complements the GWAS locus. However, the dissection of a GRN and the crosstalk underlying multiple agronomical traits, along with prioritizing important genes in eQTL-derived GRNs, remains a major challenge. In this study, we generated 558 transcriptional profiles of lint-bearing ovules at one day post-anthesis (DPA) from a selective core cotton germplasm, from which we identified 12,207 eQTLs. By integrating with a GWAS catalog, we found that 66 out of 187 (35.29%) known phenotypic GWAS loci are colocalized with 1,090 eQTLs, forming 38 major functional GRNs predominantly (30 out of 38) associated with seed size-related phenotypes. Of the eGenes, 34 were shared between at least two functional GRNs, exhibiting pleiotropic effects, such as NF-YB3, GRDP1, and IDD7. Narrow-sense heritability analysis showed that the heritability increased with combining the eQTLs with GRNs compared to those with previous yield trait GWAS loci. The extreme gradient boosting (XGBoost) machine learning approach was then applied to predict seed cotton yield phenotypes based on gene expression. Top-ranking eGenes (NF-YB3, FLA2, and GRDP1) derived by XGBoost with pleiotropic effects on yield traits were validated, along with their potential roles by correlation analysis, domestication selection analysis, and transgenic plants. This study provides insights into the mining of GRNs in relation to the pleiotropy of phenotype. The combination of eQTL and machine learning approaches is efficient in improving the genetic dissection of agricultural traits.

genetics↗

bioRxiv2022

Molecular and functional characterization of the Drosophila melanogaster conserved smORFome

Short polypeptides encoded by small open reading frames (smORFs) are ubiquitously found in eukaryotic genomes and are important regulators of physiology, development, and mitochondrial processes. Here, we focus on a subset of 298 smORFs that are evolutionarily conserved between Drosophila melanogaster and humans. Many of these smORFs are conserved broadly in the bilaterian lineage, with [~]182 conserved in plants. Within these conserved smORFs, we observed remarkably heterogenous spatial and temporal expression patterns - indicating wide-spread tissue-specific and stage-specific mitochondrial architectures. In addition, an analysis of annotated functional domains revealed a predicted enrichment of smORF polypeptides localizing to mitochondria. We conducted an embryonic ribosome profiling experiment finding support for translation of 137 of these smORFs during embryogenesis. We further embarked on functional characterization using CRISPR knockout/activation, RNAi knockdown, and cDNA overexpression, revealing diverse phenotypes. This study underscores the importance of identifying smORF function in disease and phenotypic diversity.

genomics↗

bioRxiv2021

The methylation of SDC2 and TFPI2 defined three methylator phenotypes of colorectal cancer

Methylation-based noninvasive molecular diagnostics are easy and feasible tools for the early detection of colorectal cancer (CRC). However, many of them have the limitation of low sensitivity with some CRCs detection failed in clinical practice. In this study, the clinical and pathological characteristics, as well as molecular features of three methylator-groups, defined by the promoter methylation status of SDC2 and TFPI2, were investigated in order to improve the performance of CRC detection. The Illumina Infinium 450k Human DNA methylation data and clinical information of CRCs were collected from The Cancer Genome Atlas (TCGA) project and Gene Expression Omnibus (GEO) database. CRC samples were divided into three groups, HH (dual-positive), HL (single positive) and LL (dual-negative) according to the methylation status of SDC2 and TFPI2 promoters. Differences in age, tumor location, microsatellite instable status and differentially expressed genes (DEGs) were evaluated among the three groups and these findings were then confirmed in our inner CRC dataset. The combination of methylated SDC2 and TFPI2 showed a superior performance of distinguishing CRCs from normal controls than each alone. Samples of HL group were more often originated from left-side CRCs whereas very few of them were from right-side (P < 0.05). HH grouped CRCs showed a higher level of microsatellite instability and mutation load than other two groups (mean nonsynonymous mutations for HH/HL/LL: 10.55/3.91/7.02, P = 0.0055). All mutations of BRAF, one of the five typical CpG island methylator phenotype (CIMP) related genes, were found in HH group (HH/HL/LL: 51/0/0, P = 0.018). Also there was a significantly older patient age at the diagnosis in HH group. Gene expression analysis identified 37, 84 and 22 group-specific DEGs for HH, HL and LL, respectively. Functional enrichment analysis suggested that HH specific DEGs were mainly related to the regulation of transcription and other processes, while LL specific DEGs were enriched in the biological processes of extracellular matrix interaction and cell migration. The three defined mathylator groups showed great difference in tumor location, patient age, MSI and ECM biological process, which could facilitate the development of more effective biomarkers for CRC detection.

cancer biology↗

bioRxiv2018

Large scale changes in host methylation patterns induced by IncA/C plasmid transformation in Vibrio cholerae

DNA methylation is a central epigenetic modification and has diverse biological functions in eukaryotic and prokaryotic organisms alike. The IncA/C plasmid genomes are approximately 150kb in length and harbour three methylase genes, two of which demonstrate cytosine specificity. Transformation of the Vibrio cholerae strain C6706 with the IncA/C plasmid pVC211 resulted in a significant relabelling of the methylation patterns on the host chromosomes. The new methylation patterns induced by transformation with IncA/C plasmid were accepted by the restriction enzymes of the hosts restriction modification (RM) system. These data uncover a novel mechanism by which plasmids can be compatible with a hosts RM system and suggest a possible reason that plasmids of the IncA/C family are broad-host-range.\n\nAuthor summaryAntibiotic resistance of bacteria is a growing serious problem worldwidely and the horizontal transfer of multi-drug resistance genes mediated by plasmids within and between species of bacteria is the main reason. In the researches of multi-drug resistance of Vibrio cholerae, I have isolated several IncA/C plasmids. What impressed me most is their ability to accumulate the resistant genes. Moreover, they can transfer with high frequency and are stable in several bacterial species. There are at least three Tra regions on the IncA/C plasmid which containing components of the Type 4 Secretion System and are important for conjugative transfer of plasmids. So the horizontal transfer ability of IncA/C plasmids is reasonable. There are three methylase genes on the small genome of IncA/C plasmids, which demonstrate cytosine specificity and are seldom in bacteria. Their modification target and roles are interesting. Here, we analysed the methylation profiles of the host V. choerae induced by the plasmid pVC211 and found that they were completely changed. In addition to replicons, this may be a novel mechanism that plasmid cross the barrier of the hosts RM system and become broad-host range. Changing the activity of methylase in IncA/C plasmids may be a new way to affect the stability of IncA/C plasmids to eliminate these multidrug-resistant plasmids from bacteria.

microbiology↗

bioRxiv2018

Low nutrient levels reduce the fitness cost of MexCD-OprJ efflux pump overexpression in ciprofloxacin-resistant Pseudomonas aeruginosa

The long-term persistence of antibiotic resistance in the environment is a public health concern. Expression of an efflux pump, an important mechanism of resistance to antibiotics, is usually associated with a fitness cost in bacteria. In this study, we aimed to determine why antibiotic resistance conferred by overexpression of an efflux pump persists in environments such as drinking and source water in which antibiotic selective pressure may be very low or even absent. Competition experiments between wild-type Pseudomonas aeruginosa and ciprofloxacin-resistant mutants revealed that the fitness cost of ciprofloxacin resistance (strains cip_1, cip_2, and cip_3) significantly decreased (P < 0.05) under low-nutrient (0.5 mg/l total organic carbon (TOC)) relative to high-nutrient (500 mg/l TOC) conditions. Mechanisms underlying this fitness cost were analyzed. MexD gene expression in resistant bacteria (cip_3 strain) was significantly lower (P < 0.05) in low-nutrient conditions, with 10 mg/l TOC (8.01 {+/-} 0.82-fold), than in high-nutrient conditions, with 500 mg/l TOC (48.89 {+/-} 4.16-fold). Moreover, rpoS gene expression in resistant bacteria (1.36 {+/-} 0.13-fold) was significantly lower (P < 0.05) than that in the wild-type strain (2.78 {+/-} 0.29-fold) under low-nutrient conditions (10 mg/l TOC), suggesting a growth advantage. Furthermore, the difference in metabolic activity between the two competing strains was significantly smaller (P < 0.05) in low-nutrient conditions (5 and 0.5 mg/l TOC). These results suggest that nutrient levels are a key factor in determining the persistence and spread of antibiotic resistance conferred by efflux pumps in the natural environment with trace amounts or no antibiotics.\n\nImportanceThe widespread of antibiotic resistance has led to an increasing concern about the environmental and public health risks. Mechanisms associated with antibiotic resistance including efflux pumps often increase bacterial fitness cost. Our study showed that the fitness cost of ciprofloxacin resistance conferred by overexpression of MexCD-OprJ efflux pump significantly decreased under low-nutrient relative to high-nutrient conditions. The significance of our research is to reveal that nutrient levels are key factor in determining the persistence of antibiotic resistance conferred by efflux pumps under conditions with trace amounts or no antibiotics, which can be mediated by some mechanisms including MexD gene expression, SOURs differences, and rpoS gene regulation.

microbiology↗