Biology subjects

Auerbach, S. S.

Publications and source records attributed to Auerbach, S. S..

4 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↗

bioRxiv2024

Completion of the DrugMatrix Toxicogenomics Database using ToxCompl

The DrugMatrix Database contains systematically generated toxicogenomics data from short-term in vivo studies for over 600 chemicals. However, most of the potential endpoints in the database are missing due to a lack of experimental measurements. We present our study on leveraging matrix factorization and machine learning methods to predict the missing values in the DrugMatrix, which includes gene expression across eight tissues on two expression platforms along with paired clinical chemistry, hematology, and histopathology measurements. One major challenge we encounter is the skewed distribution of the available measured data, in terms of both tissue sources and values. We propose a method, ToxiCompl, that applies systematic hybrid sampling guided by Bayesian optimization in conjunction with low-rank matrix factorization to recover the missing values. ToxiCompl achieves good training and validation performance from a machine learning perspective. We further conduct an in-depth validation of the predicted data from biological and toxicological perspectives with a series of analyses. These include examining the connectivity pattern of predicted gene expression responses, characterizing molecular pathway-level responses from sets of differentially expressed genes, evaluating known transcriptional biomarkers of tissue toxicity, and characterizing pre-dicted apical endpoints. Our analysis shows that the predicted differential gene expression, broadly speaking, aligns with what would be anticipated. For example, in most instances, our predicted differentially expressed gene lists offer a connectivity level comparable to that of measured data in connectivity analysis. Using Havcr1, a known transcriptional biomarker of kidney injury, we identify treatments that, based on the predicted expression data, manifest kidney toxicity in a manner that is mechanistically plausible and supported by the literature. Characterization of the predicted clinical chemistry data suggests that strong effects are relatively reliably predicted, while more subtle effects pose a greater challenge. In the case of histopathological prediction, we find a significant overprediction due to positivity bias in the measured data. Developing methods to deal with this bias is one of the areas we plan to target for future improvement. The main advantage of the ToxiCompl approach is that, in the absence of additional experimental data, it drastically extends the toxicogenomic landscape into a number of data-poor tissues, thereby allowing researchers to formulate mechanistic hypotheses about effects in tissues that have been underrepresented in the literature. All measured and predicted DrugMatrix data (i.e., gene expression, clinical chemistry, hematology, and histopathology) are available to the public through an intuitive GUI interface that allows for data retrieval, gene set analysis and high dimensional visualization of gene expression similarity (https://rstudio.niehs.nih.gov/complete_drugmatrix/).

genomics↗

bioRxiv2021

ALOHA: Aggregated local extrema splines for high-throughput dose-response analysis

Computational methods for genomic dose-response integrate dose-response modeling with bioinformatics tools to evaluate changes in molecular and cellular functions related to pathogenic processes. These methods use parametric models to describe each genes dose-response, but such models may not adequately capture expression changes. Additionally, current approaches do not consider gene co-expression networks. When assessing co-expression networks, one typically does not consider the dose-response relationship, resulting in co-regulated gene sets containing genes having different dose-response patterns. To avoid these limitations, we develop an analysis pipeline called Aggregated Local Extrema Splines for High-throughput Analysis (ALOHA), which computes individual genomic dose-response functions using a flexible class Bayesian shape constrained splines and clusters gene co-regulation based upon these fits. Using splines, we reduce information loss due to parametric lack-of-fit issues, and because we cluster on dose-response relationships, we better identify co-regulation clusters for genes that have co-expressed dose-response patterns from chemical exposure. The clustered pathways can then be used to estimate a dose associated with a pre-specified biological response, i.e., the benchmark dose (BMD), and approximate a point of departure dose corresponding to minimal adverse response in the whole tissue/organism. We compare our approach to current parametric methods and our biologically enriched gene sets to cluster on normalized expression data. Using this methodology, we can more effectively extract the underlying structure leading to more cohesive estimates of gene set potency.

bioinformatics↗

bioRxiv2019

The sensitivity of transcriptomics BMD modeling to the methods used for microarray data normalization

Whole-genome expression data generated by microarray studies have shown promise for quantitative human health risk assessment. While numerous approaches have been developed to determine benchmark doses (BMDs) from probeset-level dose responses, sensitivity of the results to methods used for normalization of the data has not yet been systematically investigated. Normalization of microarray data converts raw hybridization signals to expression estimates that are expected to be proportional to the amounts of transcripts in the profiled specimens. Different approaches to normalization have been shown to greatly influence the results of some downstream analyses, including biological interpretation. In this study we evaluate the influence of microarray normalization methods on the transcriptomic BMDs. We demonstrate using in vivo data that the use of alternative pipelines for normalization of Affymetrix microarray data can have a considerable impact on the number of detected differentially expressed genes and pathways (processes) determined to be treatment responsive, which may lead to alternative interpretations of the data. In addition, we found that normalization can have a considerable effect (as much as [~]30-fold in this study) on estimation of the minimum biological potency (transcriptomic point of departure). We argue for consideration of alternative normalization methods and their data-informed selection to most effectively interpret microarray data for use in human health risk assessment.

pharmacology and toxicology↗