Biology subjects

Madad, S.

Publications and source records attributed to Madad, S..

3 recordsLinked to original sources

bioRxiv2026

MetagenomicsBench: A Verifiable Benchmark for Agentic Metagenomics Analysis

Metagenomics has transformed our ability to characterize microbial communities without relying on cultivation, but translating complex microbiome datasets into reliable biological conclusions remains a major analytical challenge. Although AI agents have improved substantially at software engineering and general data analysis, it remains unclear whether they can reliably analyze and interpret real-world metagenomic data. We introduce MetagenomicsBench, a benchmark of 100 verifiable evaluations derived from published datasets spanning community structure, host-microbiome associations, microbial function, longitudinal dynamics, and microbiome interventions. Each evaluation provides experimental data and a deterministic grader that evaluates recovery of a key analytical or biological result. Across 10,200 trajectories from 34 model-harness configurations, the strongest configuration achieved a 60.0% pass rate. Performance varied across functional domains, data modalities, and biological systems, while greater cost, token usage, and tool use did not consistently correspond to higher accuracy. Trajectory analysis showed that agents often produced technically plausible analyses while making mistakes in problem interpretation, statistical reasoning, and biological interpretation. MetagenomicsBench provides a framework for measuring progress toward AI agents that can reliably make and validate the analytical and biological decisions required for metagenomic research.

microbiology↗

bioRxiv2026

An integrated single cell and spatial omics atlas of human prenatal development

Single cell genomics has enabled analysis of human prenatal development at unprecedented resolution. However, most studies have relied on dissociated tissues during restricted windows of development, limiting insights into how spatially distributed networks of cells, and multicellular niches emerge and adapt to distinct organ microenvironments in situ. Moreover, existing human developmental atlases have not yet been harmonised, and we thus lack a comprehensive catalogue of known cell types in the developing human body. Here, we introduce the Human Developmental Cell Atlas (HDCA), a unified structural, cellular and molecular resource for prenatal human development. The HDCA integrates published and unpublished single cell/nucleus RNAseq atlases across prenatal organs, and includes a newly generated, spatially resolved, multimodal cell atlas of intact human embryos. Spanning 4-22 post conceptional weeks, capturing embryonic and early to mid fetal stages, the HDCA contains [~]4.6 million cells/nuclei which resolve into [~]450 cell types, explorable with a bespoke web portal. For a global overview of the human embryos multicellular communities, we applied unsupervised deep learning to our intact human embryo spatial data, charting 114 tissue niches that are structural and signalling hubs for the cellular interactions of the embryo. Guided by these niches, we profiled cellular networks over space and time, not examinable using single-organ atlases. In so doing, we revealed tissue-specific fibroblast patterning from previously undescribed mesenchyme progenitors, early diversification of organ-specific blood capillaries and lymphatic vasculature, emergence of neural crest cell fates, the formation of placode-and neural crest-derived peripheral sensory neurons, and how tissue niches guide peripheral neuron maturation and axonal migration. The HDCA thus serves as a comprehensive step towards a comprehensive understanding of human prenatal development, and a template towards unravelling the biology of congenital disorders.

developmental biology↗

bioRxiv2022

Dictionary learning for integrative, multimodal, and scalable single-cell analysis

Mapping single-cell sequencing profiles to comprehensive reference datasets represents a powerful alternative to unsupervised analysis. Reference datasets, however, are predominantly constructed from single-cell RNA-seq data, and cannot be used to annotate datasets that do not measure gene expression. Here we introduce bridge integration, a method to harmonize singlecell datasets across modalities by leveraging a multi-omic dataset as a molecular bridge. Each cell in the multi-omic dataset comprises an element in a dictionary, which can be used to reconstruct unimodal datasets and transform them into a shared space. We demonstrate that our procedure can accurately harmonize transcriptomic data with independent single cell measurements of chromatin accessibility, histone modifications, DNA methylation, and protein levels. Moreover, we demonstrate how dictionary learning can be combined with sketching techniques to substantially improve computational scalability, and harmonize 8.6 million human immune cell profiles from sequencing and mass cytometry experiments. Our approach aims to broaden the utility of single-cell reference datasets and facilitate comparisons across diverse molecular modalities. AvailabilityInstallation instructions, documentations, and vignettes are available at http://www.satijalab.org/seurat

genomics↗