Biology subjects

Motsinger-Reif, A.

Publications and source records attributed to Motsinger-Reif, A..

7 recordsLinked to original sources

bioRxiv2026

Tox21mer, A transformer foundation model for Tox21 high-throughput concentration-response curves data

The U.S. Tox21 collaboration has generated a large reference library of high-throughput concentration-response assays. Here we present Tox21mer, a 43.5-million-parameter transformer that encodes each Tox21 concentration-response curve together with assay metadata into a 768-dimensional representation. Tox21mer was pretrained on ~2.5 million curves from 102 assay protocols and 6,727 compounds using masked-response reconstruction as the primary objective, with low-weight auxiliary supervision on assay outcome and AC50. To evaluate the learned representation, we trained lightweight probes on frozen embeddings from concentration-response curves of held-out compounds. The representation supported a macro-F1 of 0.985 for three-class outcome prediction (agonist, antagonist, inactive), a binary F1 of 0.994 for active/inactive prediction, and an R2 of 0.87 for log10(AC50). The learned embeddings formed coherent groupings by curve-class category. A masked-only pretraining variant retained near-baseline probe performance, indicating that the representation is learned largely from the self-supervised objective rather than from auxiliary labels. Ablation analyses further showed that predictive performance depends mainly on curve-level response-value distributions conditioned on assay context, with limited reliance on detailed within-curve ordering. Tox21mer thus provides a reusable foundation representation for Tox21 concentration-response data that can support extrapolation to untested compounds through integration with chemical features or distillation into chemistry-only student models for large-scale external screening.

bioinformatics↗

bioRxiv2025

Mapping of genotype-by-environment interaction loci for Metabolic Syndrome-like traits using the multi-parent Drosophila Synthetic Population Resource determines that main genetic effects are distinct from environment dependent plastic loci

Metabolic Syndrome (MetS) risk, driven by genotype-environment interactions like diet, is rising globally. Due to its genetic and environmental complexity, the genetic architecture and interconnected traits underlying MetS is poorly understood. In Drosophila, genotype-by-diet interactions significantly influence MetS-like traits. This study used the Drosophila Synthetic Population Resource to dissect the genetic architecture of both genotypic and genotype-by-diet interaction effects underlying trait variation. The study hypotheses were: 1) Loci responsible for metabolic phenotypic variation should be shared across traits. 2) Genetic loci responsible for plasticity and epistatic interactions for metabolic traits should also be the loci responsible for the main effects. 3) Genes responsible for variation in metabolic traits should share common functions. Using a round-robin crossing scheme and novel analyses, we mapped additive, dominance, and epistatic loci--some diet-specific, others diet-independent. Main-effect and plastic loci were largely distinct, as were epistatic loci from main-effect loci, highlighting that main genetic effects alone will not explain how genetic variants interact with the environment or the genome to influence disease risk. gene-by-diet or gene-by-gene interactions influencing MetS risk. Further, tremendous cryptic genetic variation for metabolic traits is lurking in natural populations. We explored the function of candidate genes from our study empirically and with bioinformatics. While some of the candidate genes might have been expected, most would not have been identified a priori, thus with this study we have identified many new candidate mechanisms contributing to the genetic and genotype-by-diet interaction effects on MetS variance.

genetics↗

bioRxiv2023

TaxaNorm: a novel taxa-specific normalization approach for microbiome data

BackgroundIn high-throughput sequencing studies, sequencing depth, which quantifies the total number of reads, varies across samples. Unequal sequencing depth can obscure true biological signals of interest and prevent direct comparisons between samples. To remove variability due to differential sequencing depth, taxa counts are usually normalized before downstream analysis. However, most existing normalization methods scale counts using size factors that are sample specific but not taxa specific, which can result in over- or under-correction for some taxa. ResultsWe developed TaxaNorm, a novel normalization method based on a zero-inflated negative binomial model. This method assumes the effects of sequencing depth on mean and dispersion vary across taxa. Incorporating the zero-inflation part can better capture the nature of microbiome data. We also propose two corresponding diagnosis tests on the varying sequencing depth effect for validation. We find that TaxaNorm achieves comparable performance to existing methods in most simulation scenarios in downstream analysis and reaches a higher power for some cases. Specifically, it has a well balance on power and false discoveries control. When applying the method in a real dataset, TaxaNorm has improved performance when correcting technical bias. ConclusionTaxaNorm considers correcting both sample- and taxon-specific bias by introducing an appropriate regression framework in the microbiome data, which aids in data interpretation and visualization. The TaxaNorm R package is freely available through the CRAN repository https://CRAN.R-project.org/package=TaxaNorm and the source code can be downloaded at https://github.com/wangziyue57/TaxaNorm.

bioinformatics↗

bioRxiv2019

Untargeted Metabolomic Profiling Identifies Disease-specific Profiles in Food Allergy

BackgroundFood allergy (FA) affects an increasing proportion of children for reasons that remain obscure. Identification of pathogenic mechanisms involved in FA using untargeted metabolomic approaches may provide much needed diagnostic and prognostic disease biomarkers and improved treatment options.\n\nMethodsMass spectrometry-based untargeted metabolomic profiling was performed on serum samples of children with either FA alone, asthma alone or both FA and asthma as well as healthy pediatric controls.\n\nResultsFA subjects exhibited a disease-specific metabolomic signature as compared to both control subjects and asthmatics. In particular, FA was uniquely associated with a marked decrease in sphingolipids, as well as a number of other lipid metabolites, in the face of normal frequencies of circulating natural killer T (NKT) cells. Specific comparison of FA and asthmatic subjects revealed differences in the microbiota-sensitive aromatic amino acid and secondary bile acid metabolism. Children with both FA and asthma exhibited a metabolomic profile that aligned with that of FA alone but not asthma. Among children with FA, distinct profiles were associated with history of severe reactions and presence of multiple FA.\n\nConclusionsChildren with FA display a disease-specific metabolomic profile that is informative of disease mechanisms and severity, and which dominates in the presence of asthma. Lower levels of sphingolipids and ceramides and other metabolomic alterations observed in FA children may reflect the interplay between an altered microbiota and immune cell subsets in the gut.

immunology↗

bioRxiv2019

Same-Species Contamination Detection with Variant Calling Information from Next Generation Sequencing

MotivationSame-species contamination detection is an important quality control step in genetic data analysis. Compared with widely discussed cross-species contamination, same-species contamination is more challenging to detect, and there is a scarcity of methods to detect and correct for this quality control issue. Same-species contamination may be due to contamination by lab technicians or samples from other contributors. Here, we introduce a novel machine learning algorithm to detect same species contamination in next generation sequence data using support vector machines. Our approach uniquely detects such contamination using variant calling information stored in the variant call format (VCF) files (either DNA or RNA), and importantly can differentiate between same species contamination and mixtures of tumor and normal cells. MethodsIn the first stage of our approach, a change-point detection method is used to identify copy number variations or copy number aberrations (CNVs or CNAs) for filtering prior to testing for contamination. Next, single nucleotide polymorphism (SNP) data is used to test for same species contamination using a support vector machine model. Based on the assumption that alternative allele frequencies in next generation sequencing follow the beta-binomial distribution, the deviation parameter {rho} is estimated by maximum likelihood method. All features of a radial basis function (RBF) kernel support vector machine (SVM) are generated using either publicly available or private training data. Lastly, the generated SVM is applied in the test data to detect contamination. If training data is not available, a default RBF kernel SVM model is used. ResultsWe demonstrate the potential of our approach using simulation experiments, creating datasets with varying levels of contamination. The datasets combine, in silico, exome sequencing data of DNA from two lymphoblastoid cell lines (NA12878 and NA10855). We generated VCF files using variants identified in these data, and then evaluated the power and false positive rate of our approach to detect same species contamination. Our simulation experiments show that our method can detect levels of contamination as low as 5% with reasonable false positive rates. Results in real data have sensitivity above 99.99% and specificity at 90.24%, even in the presence of DNA degradation that has similar features to contaminated samples. Additionally, the approach can identify the difference between mixture of tumor-normal cells and contamination. We provide an R software implementation of our approach using the defcon()function in the vanquish: Variant Quality Investigation Helper R package on CRAN.

bioinformatics↗

bioRxiv2018

Identifying individual risk rare variants using protein structure-guided local tests (POINT)

Rare variants are of increasing interest to genetic association studies because of their etiological contributions to human complex diseases. Due to the rarity of the mutant events, rare variants are routinely analyzed on an aggregate level. While aggregation analyses improve the detection of global-level signal, they are not able to pinpoint causal variants within a variant set. To perform inference on a localized level, additional information, e.g., biological annotation, is often needed to boost the information content of a rare variant. Following the observation that important variants are likely to cluster together on functional domains, we propose a protein structure guided local test (POINT) to provide variant-specific association information using structure-guided aggregation of signal. Constructed under a kernel machine framework, POINT performs local association testing by borrowing information from neighboring variants in the 3-dimensional protein space in a data-adaptive fashion. Besides merely providing a list of promising variants, POINT assigns each variant a p-value to permit variant ranking and prioritization. We assess the selection performance of POINT using simulations and illustrate how it can be used to prioritize individual rare variants in PCSK9 associated with low-density lipoprotein in the Action to Control Cardiovascular Risk in Diabetes (ACCORD) clinical trial data.\n\nAuthor summaryWhile it is known that rare variants play an important role in understanding associations between genotype and complex diseases, pinpointing individual rare variants likely to be responsible for association is still a daunting task. Due to their low frequency in the population and reduced signal, localizing causal rare variants often requires additional information, such as type of DNA change or location of variant along the sequence, to be incorporated in a biologically meaningful fashion that does not overpower the genotype data. In this paper, we use the observation that important variants tend to cluster together on functional domains to propose a new approach for prioritizing rare variants: the protein structure guided local test (POINT). POINT uses a genes 3-dimensional protein folding structure to guide aggregation of information from neighboring variants in the protein in a robust manner. We show how POINT improves selection performance over single variant tests and sliding window approaches. We further illustrate how it can be used to prioritize individual rare variants using the Action to Control Cardiovascular Risk in Diabetes (ACCORD) clinical trial data, finding five promising variants within PCSK9 in association with low-density lipoprotein, including three new mutations near the PCSK9-LDLR binding domain.

bioinformatics↗

bioRxiv2018

Altered Bile Acid Profile Associates with Cognitive Impairment in Alzheimer’s Disease – An Emerging Role for Gut Microbiome

IntroductionIncreasing evidence suggests a role for the gut microbiome in central nervous system disorders and specific role for the gut-brain axis in neurodegeneration. Bile acids (BA), products of cholesterol metabolism and clearance, are produced in the liver and are further metabolized by gut bacteria. They have major regulatory and signaling functions and seem dysregulated in Alzheimer disease (AD).\n\nMethodsSerum levels of 15 primary and secondary BAs and their conjugated forms were measured in 1,464 subjects including 370 cognitively normal older adults (CN), 284 with early mild cognitive impairment (MCI), 505 with late MCI, and 305 AD cases enrolled in the AD Neuroimaging Initiative. We assessed associations of BA profiles including selected ratios with diagnosis, cognition, and AD-related genetic variants, adjusting for cofounders and multiple testing.\n\nResultsIn AD compared to CN, we observed significantly lower serum concentrations of a primary BA (cholic acid CA) and increased levels of the bacterially produced, secondary BA, deoxycholic acid (DCA), and its glycine and taurine conjugated forms. An increased ratio of DCA:CA, which reflects 7-dehydroxylation of CA by gut bacteria, strongly associated with cognitive decline, a finding replicated in serum and brain samples in the Rush Religious Orders and Memory and Aging Project. Several genetic variants in immune response related genes implicated in AD showed associations with BA profiles.\n\nConclusionWe report for the first time an association between altered BA profile, genetic variants implicated in AD and cognitive changes in disease using a large multicenter study. These findings warrant further investigation of gut dysbiosis and possible role of gut liver brain axis in the pathogenesis of AD.

neuroscience↗