bioRxiv Science⌕ Search

SEARCH · bioRxiv Science

Results for “Preprint”

Search indexed bioRxiv preprints in genomics, neuroscience, cell biology and bioinformatics. Read source abstracts and check manuscript versions; preprints are not peer reviewed.

Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.

At least 289 records · Page 16Linked to original sources
bioRxiv2019

Genome-wide association study of cryptosporidiosis in infants implicates PRKCA

Diarrhea is a major cause of both morbidity and mortality worldwide, especially among young children. Cryptosporidiosis is a leading cause of diarrhea in children, particularly in South Asia and Sub-Saharan Africa where it is responsible for over 200,000 deaths per year. Beyond the initial clinical presentation of diarrhea, it is associated with long term sequelae such as malnutrition and neurocognitive developmental deficits. Risk factors include poverty and overcrowding, yet not all children with these risk factors and exposure are infected, nor do all infected children develop symptomatic disease. One potential risk factor to explain these differences is their human genome. To identify genetic variants associated with symptomatic cryptosporidiosis, we conducted a genome-wide association study (GWAS) examining 6.5 million single nucleotide polymorphisms (SNPs) in 873 children from three independent cohorts in Dhaka, Bangladesh: the Dhaka Birth Cohort (DBC), the Performance of Rotavirus and Oral Polio Vaccines in Developing Countries (PROVIDE) study, and the Cryptosporidiosis Birth Cohort (CBC). Associations were estimated separately for each cohort under an additive model, adjusting for height-for-age Z-score at 12 months of age, the first two principal components to account for population substructure, and genotyping batch. The strongest meta-analytic association was with rs58296998 (P=3.73x10-8), an intronic SNP and eQTL of PRKCA. Each additional risk allele conferred 2.4 times the odds of cryptosporidiosis in the first year of life. This genetic association suggests a role for protein kinase C alpha in pediatric cryptosporidiosis and warrants further investigation. This article was submitted to an online preprint archive.(1)\n\nImportanceGlobally, one of the major causes of pediatric morbidity and mortality remains diarrhea. The initial symptoms of diarrhea can often lead to long term consequences for the health of young children, such as malnutrition and neurocognitive developmental deficits. Despite many children having similar exposures to infectious causes of diarrhea, not all develop symptomatic disease, indicating a possible role for human genetic variation. Here we conducted a genetic study of susceptibility to symptomatic disease associated with Cryptosporidium infection (a leading cause of diarrhea) in three independent cohorts of infants from Dhaka, Bangladesh. We identified a genetic variant within protein kinase c alpha (PRKCA) associated with higher risk of cryptosporidiosis in the first year of life. These results indicate a role for human genetics in susceptibility to cryptosporidiosis and warrant further research to elucidate the mechanism.

genetics↗
bioRxiv2019

Differential robustness to specific potassium channel deletions in midbrain dopaminergic neurons

The authors have withdrawn this preprint titled "Differential robustness to specific potassium channel deletions in midbrain dopaminergic neurons". Upon review of breeding and genotyping data, Kcnn3-/- mice could not be trusted as representative of the expected genetic deletion. As a consequence data generated from these animals do not constitute a valid description of the Kcnn3-/- genotype in dopaminergic neurons. Therefore, the authors do not wish this work to be cited as reference for the project. If you have any questions, please contact the corresponding author.

neuroscience↗
bioRxiv2019

PAX8 expressing epithelial cells are a cancer-prone source of clonal cyclical regeneration of endometrial epithelium

bioRxiv has withdrawn this manuscript after a formal investigation by Cornell University that concluded that it was submitted and made public without listing all contributing authors. The list of authors in this preprint should have been: Dah-Jiun Fu, Andrea J. De Micheli, Blaine A. Harlan, Mallikarjun Bidarimath, Minseok Kim, Lora H. Ellenson, Benjamin D. Cosgrove, Andrea Flesken-Nikitin and Alexander Yu. Nikitin, with DJF, AJDM and BAH indicated as equal contributors. Therefore, this work should not be cited as reference for the project. If you have any questions, please contact the corresponding authors.

cancer biology↗
bioRxiv2025

Funders' expectations for open science in cardiovascular research: A Scoping review of the largest cardiovascular research funders

Open science practices, including data sharing, open access, and prospective study registration, have been increasingly recognized to improve transparency, reproducibility, and accessibility in research, yet their uptake and implementation by cardiovascular research funders is unclear. We conducted a scoping review of publicly available policies, guidance, and grant instructions from 12 members of the Global Cardiovascular Research Funders Forum to assess expectations, monitoring, and support for open science in cardiovascular research. We included 105 documents from 9 funders; no relevant documents were identified for 3 funders. Data sharing (75%) and open access (67%) were the most common mandates by funders, followed by prospective registration (50%). Requirements for other practices, including code sharing, use of reporting guidelines, preprints, and open peer review, were uncommon. Monitoring compliance was inconsistent, with many funders not specifying any mechanisms, even for widely required practices. Where available, support was most often provided through financial assistance, guidance, or infrastructure, particularly for open access, data sharing, and patient or public involvement. These findings suggest that while cardiovascular funders are engaging with open science, policies remain uneven in scope, monitoring, and support. Navigating the open science implementation gap in cardiovascular research will be essential to reap the benefits of transparency and innovation, only possible through the sharing of information and data.

scientific communication and education↗
bioRxiv2025

The Lsm1-7 complex marginally affects unstable mRNA degradation in yeast

Withdrawal StatementThe authors have withdrawn this manuscript because additional controls did not confirm the involvement of LSM1 in NMD. Our initial conclusions were based on wrong assumptions about the used degron system and the required controls. RNA sequencing, phenotype and RT-qPCR results are affected, but other results presented in the manuscript remain valid. We apologize to the colleagues who based any of their work or conclusion on that part of the preprint. Therefore, the authors do not wish this work to be cited as reference for the project. If you have any questions, please contact the corresponding author.

molecular biology↗
bioRxiv2026

Plasmodium falciparum stomatin-like protein forms a putative complex with a metalloprotease in distinct mitochondrial loci

Withdrawal statementThis manuscript has been withdrawn owing to a duplicate posting of manuscript number BIORXIV/2024/604071. This was a technical error for which the authors are not responsible, and this DOI for the work should not be cited as reference. The record is retained here for purposes of transparency. If you have any questions, please contact the corresponding author. The correct preprint can be found at doi: 10.1101/2024.07.18.604071

microbiology↗
bioRxiv2026

Non-viral in vivo electroporation-based chromosomal engineering and repair assessment in the murine uterine epithelium

Withdrawal StatementThis manuscript has been withdrawn owing to a duplicate posting of manuscript number BIORXIV/2025/640089. This was a technical error for which the authors are not responsible, and this DOI for the work should not be cited as reference. The record is retained here for purposes of transparency. If you have any questions, please contact the corresponding author. The correct preprint can be found at doi: 10.1101/2025.02.25.640089.

genomics↗
bioRxiv2026

A novel protocol for the efficient generation of all three major hippocampal neuronal sub-populations from human pluripotent stem cells

The diverse computational functions of the human hippocampus rely on coordinated interactions among dentate gyrus (DG), CA3, and CA1 subfields, yet generating all three neuronal identities in vitro - particularly CA1 - has remained challenging. Here we establish a reproducible and modular differentiation protocol that directs human pluripotent stem cells (hPSCs) through dorsomedial telencephalic progenitors to yield DG, CA3, and CA1 neuronal subtypes together with hippocampal regionally specified astrocytes. Early tri-inhibition combined with Sonic hedgehog suppression produced dorsal forebrain progenitors (FOXG1+, PAX6+), while FGF2 treatment supported progenitor maintenance and induced TBR2+ intermediate progenitors. Controlled WNT activation using CHIR99021 drove progressive enrichment of PROX1 hippocampal progenitors across two independent donor lines. Terminal differentiation produced MAP2+/TAU+ neurons that expressed DG (PROX1), CA3 (GRIK4), and CA1 (WFS1, OCT6) markers, with maturing synaptic puncta. Defined progenitors generated long-lived (>400 days) hippocampal organoids exhibiting mixed neuronal-glial populations and spontaneous activity characterized by increased firing rates, high information entropy, and hub-like causal connectivity relative to monolayers, whereas astrocytes-supplemented monolayers displayed intermediate maturation. Population level electrophysiological analysis was also conducted to explore the dynamics of these different cultures. This platform enables systematic experimental control over neuron-astrocyte ratios, culture geometry, and developmental timing, providing a foundation for mechanistic studies of human hippocampal development, circuit function, and disease. Note on figure qualityThis is the preprint version of the manuscript. Figures are included adjacent to described results for the convenience of the reader but may be lower resolution due to file size restrictions on bioRxiv. High resolution figures are included as separate .tiff files for download.

neuroscience↗
bioRxiv2026

Democratised single cell proteomics uncovers tumor associated macrophage signature in CYLD cutaneous syndrome skin tumors

Single-cell proteomics (SCP) reveals cellular heterogeneity and biological insights inaccessible to bulk analysis. Existing limitations are cost, sample loss during processing, and accessibility to state-of-the-art instrumentation. We describe a label free SCP methodology in human tissue, combining fluorescence activated cell sorting (FACS), oil-immersion cell handling, mass-spectrometry, and neural-network derived spectral libraries which address these issues. We tested this methodology in a skin tumor syndrome, CYLD cutaneous syndrome (CCS), assessing tumor heterogeneity. Using a Bruker timsTOF HT platform we quantified > 4000 proteins, averaging [~]700 per cell, through a cost-effective pipeline without specialised liquid handling infrastructure. By utilising pre-existing bioinformatic tools from the scRNA-seq field we implemented a robust analysis methodology, discriminating between macrophages, dendritic cells and tumor keratinocytes, in an unbiased analysis of 419 CCS tumor cells. We validated the biological accuracy of cell annotations by cross referencing with each cells FACS markers. Furthermore, we identified a novel CCS tumor associated macrophage population which carried a tumor microenvironment remodelling signature. Our findings demonstrate an accessible SCP technology capable of yielding novel biological discoveries in clinical tissue. Graphical abstract O_FIG O_LINKSMALLFIG WIDTH=200 HEIGHT=86 SRC="FIGDIR/small/704354v1_ufig1.gif" ALT="Figure 1"> View larger version (33K): org.highwire.dtl.DTLVardef@6fe6a3org.highwire.dtl.DTLVardef@8f0ff9org.highwire.dtl.DTLVardef@61e037org.highwire.dtl.DTLVardef@1c76a43_HPS_FORMAT_FIGEXP M_FIG C_FIG IMPORTANTO_LIManuscripts submitted to Review Commons are peer reviewed in a journal-agnostic way. C_LIO_LIUpon transfer of the peer reviewed preprint to a journal, the referee reports will be available in full to the handling editor. C_LIO_LIThe identity of the referees will NOT be communicated to the authors unless the reviewers choose to sign their report. C_LIO_LIThe identity of the referee will be confidentially disclosed to any affiliate journals to which the manuscript is transferred. C_LI GUIDELINESO_LIFor reviewers: https://www.reviewcommons.org/reviewers C_LIO_LIFor authors: https://www.reviewcommons.org/authors C_LI CONTACTThe Review Commons office can be contacted directly at: office@reviewcommons.org

biochemistry↗
bioRxiv2026

Token Alignment for Verifying LLM-Extracted Text

Large language models excel at text extraction, but they sometimes hallucinate. A simple way to avoid hallucinations is to remove any extracted text that does not appear in the original source. This is easy when the extracted text is contiguous (findable with exact string matching), but much harder when it is discontiguous. Techniques for finding discontiguous phrases depend heavily on how the text is split--i.e., how it is tokenized. In this study, we show that splitting text along subword boundaries, with LLM-specific tokenizers, and aligning extracted text with ordered alignment algorithms, improves alignment by about 50% compared to word-level tokenization. To demonstrate this, we introduce the Berkeley Ordered Alignment of Text (BOAT) dataset, a modification of the Stanford Question Answering Dataset (SQuAD) that includes non-contiguous phrases, and BIO-BOAT a biomedical variant built from 51 bioRxiv preprints. We show that text-alignment methods form a partially ordered set, and that ordered alignment is the most practical choice for verifying LLM-extracted text. We implement this approach in taln, which enumerates ordinal subword alignments.

bioinformatics↗
bioRxiv2026

Targeting CDK12/CYCLIN K induces HIV gene activation and latency reversal which is mediated by P-TEFb

The administration of antiretroviral therapy has successfully suppressed the replication of Human immunodeficiency virus and substantially inhibited viral infection. However, HIV still persists in long-lived infected cell reservoirs that are resistant to therapy and to immune clearance, therefore a barrier for elimination of infection. HIV gene expression is tightly regulated by the cellular transcription machinery, where low levels of host and viral transcription factors and epigenetic constrains maintain viral persistence. Here, we show that selective targeting of CDK12/CYCLIN K (CCNK) induces both HIV-specific and a global gene activation program, which is mediated by positive transcription elongation factor b (P-TEFb). Targeting CDK12/CCNK triggers a robust HIV gene expression, accompanied with latency reversal and synergistic reactivation when combined with latency-reversing agents. Mechanistically, CDK12/CCNK inhibition promotes the recruitment of RNA Polymerase II and CDK9 to the viral promoter, accompanied with elevated levels of histone activation makers. Targeting CDK12/ CCNK also exerts a global gene activation program, by releasing P-TEFb from 7SK snRNP, and reshaping the chromatin landscape at promoter-proximal loci and gene bodies. Together, these results uncover a previously unrecognized compensatory interplay between transcriptional kinases that rewires the cellular gene expression program with implications for HIV latency reversal. IMPORTANTO_LIManuscripts submitted to Review Commons are peer reviewed in a journal-agnostic way. C_LIO_LIUpon transfer of the peer reviewed preprint to a journal, the referee reports will be available in full to the handling editor. C_LIO_LIThe identity of the referees will NOT be communicated to the authors unless the reviewers choose to sign their report. C_LIO_LIThe identity of the referee will be confidentially disclosed to any affiliate journals to which the manuscript is transferred. C_LI GUIDELINESO_LIFor reviewers: https://www.reviewcommons.org/reviewers C_LIO_LIFor authors: https://www.reviewcommons.org/authors C_LI CONTACTThe Review Commons office can be contacted directly at: office@reviewcommons.org

microbiology↗
bioRxiv2026

Automated localization of calling birds with small passive acoustic arrays in complex soundscapes

Accurately localizing vocalizing animals from passive acoustic recordings remains challenging in complex outdoor soundscapes. Although automated detection and classification of bird calls have advanced rapidly, reliable spatial localization has lagged, particularly for small arrays of autonomous recorders operating without manual intervention. Here we describe a fully automated pipeline for three-dimensional localization of bird vocalizations using distributed networks of 4 to 6 GPS-synchronized recorders deployed in heterogeneous forest environments. Our framework integrates automated multi-recorder event matching, frequency-selective FFT-based cross-correlation for time-difference-of-arrival (TDOA) estimation, geometric cycle-consistency filtering to resolve ambiguous cross-correlation peaks, and nonlinear optimization of source location and effective sound speed. Applied to multi-year datasets from three field sites, the localizations exhibit strong concordance of localizations with independently known landscape features and species-specific habitat associations. These analyses indicate that small, practical arrays can recover ecologically meaningful spatial structure in complex soundscapes without manual curation. This preprint documents the current state of the system and its performance under realistic field conditions.

ecology↗
bioRxiv2026

Macrophage Iron Metabolism in Allografts and Tumors

Macrophages are present at high frequencies in transplanted organs and solid tumors. These infiltrating macrophages are highly tuned by tissue niche signals. However, the underlying mechanisms that orchestrate the activation of distinct macrophage populations remain unclear. Here, using heart transplantation and tumor implantation models, we found that macrophages in cardiac allografts exhibited higher intracellular iron (Fe2+) level and higher expression of the iron transporter SLC11A1 than those in the solid tumors. In mice, the myeloid-specific deletion of Slc11a1 alleviated the proinflammatory state of macrophages and cardiac allografts rejection. Together, our findings identify SLC11A1 as a new player in immune response regulation, implicating SLC11A1 as a therapeutic target for both allograft and tumor rejection. This preprint reports initial findings; additional experiments are ongoing and will be included in a future full manuscript.

immunology↗
bioRxiv2026

STAR Suite: Integrating transcriptomics through AI software engineering in the NIH MorPhiC consortium

Background. STAR is the aligner underlying most transcriptomics processing, and is used inside Cell Ranger. However, the surrounding steps - adapter trimming, sorting, feature-barcode assignment, probe handling, SLAM-seq analysis, quantification, and quality control - run as separate tools scripted around it. Cell Ranger hides this complexity but is proprietary, restricted to 10x products, and barred by its license from redistribution, so it cannot serve as a shared processing layer. While STARsolo supports scRNA-seq, it lacks the feature-barcode detection needed for Perturb-seq, where short tags such as CRISPR guides or lineage barcodes are read alongside the transcriptome. No open-source tool processes Perturb-seq feature barcodes at production scale, and the only open-source processor for 10x Flex - a probe-based assay for fixed and FFPE tissue that multiplexes several samples in one lane - is a recent standalone k-mer tool. Results. STAR Suite compiles these steps into the STAR codebase to create a drop-in executable for bulk RNA-seq, scRNA-seq, Perturb-seq, 10x Flex, and SLAM-seq. It adds no dependencies beyond OpenSSL's libcrypto, and legacy behavior is preserved. On consortium and public benchmarks it is up to 41-fold faster than Cell Ranger 9.0.1 for Flex (20- to 41-fold from binary CBQ input, 17- to 23-fold from FASTQ) and 3.9- to 6.2-fold faster for scRNA-seq and Perturb-seq, timed on one modest server (24 cores, 128 GiB, local SSD) except for the largest dataset, which ran on a cloud instance limited to the same size. The concordance to Cell Ranger is high: gene-level Spearman and Pearson 0.98-1.0, cell-level Pearson on per-cell total counts 0.9999-1.0, feature-UMI Pearson 0.999-1.000, Jaccard agreement of the called-cell sets 0.97-0.995, and 99-100% CRISPR-call agreement. It also needs far less disk: its peak use on the largest Flex dataset is 16.9 GiB, against 1.8 TiB for Cell Ranger. These gains arise from running the steps in one process rather than as separate programs exchanging files, and from four new computational techniques: fast-Hamming, a vectorized exact Hamming-distance search, used for Flex probe matching and optionally for feature barcodes; a disambiguated hash cascade that assigns Flex probes without alignment; a permit-based thread scheduler that interleaves alignment and feature assignment; and variance-based auto-trimming for SLAM-seq conversion calling, which finds the reliable region of each read from how variable the conversion rate is along it. The four modules add 183,606 lines of C/C++ to STAR's 28,228. STAR Suite is the NIH MorPhiC consortium's production processor. The same binary is run by people, by AI agents and by cluster schedulers, and every production run is recorded with its commands, checksums and outputs. It has been built and maintained over eight months and 25 releases under a human-directed, AI-implemented workflow, with its full design and benchmark record public. Conclusions. STAR Suite provides the first production-ready open-source implementation of Perturb-seq feature-barcode processing, and an open end-to-end 10x Flex pipeline matched to Cell Ranger, delivered as a drop-in replacement for the STAR aligner. The design also extends: downstream analysis now done in Python or R packages can be built into the same binary, and further modalities added alongside, as a companion preprint does for ATAC-seq. Source code, workflow recipes, and per-run provenance are released under the MIT license.

bioinformatics↗
bioRxiv2026

Robust PHP in Adult Hippocampus: Essential Assay Optimizations

Presynaptic homeostatic plasticity (PHP) is a potent form of homeostatic plasticity that has been documented at synapses as diverse as the glutamatergic Drosophila neuromuscular junction (NMJ), cholinergic mammalian NMJ (including human), and glutamatergic synapses in the mammalian brain. Published experimental evidence in favor of PHP in adult hippocampus and cerebellum includes patch-clamp electrophysiology, presynaptic capacitance measurement, calcium imaging, optical reporters of vesicle release and correlated three-dimensional electron microscopy. These studies are grounded in newly optimized experimental protocols that differ substantively from those typically used to study activity-dependent plasticity in neonatal and juvenile slice preparations. Here, we elaborate and extend our assays and methodologies for the study of PHP in the adult mammalian brain. Our assays are designed to optimize synapse, cell and tissue health and minimize the incorporation of unintended adverse experimental conditions that may interfere with the induction and/or expression of PHP. In addition, we provide benchmark criteria for assessment of cell health, necessary for analysis of PHP and, in so doing, advance our understanding of postsynaptic conditions necessary for PHP induction in the adult brain. Our data underscore why PHP may have been previously overlooked, inclusive of a recent manuscript challenging the robust expression of PHP in the mammalian brain (Dou et al., 2026 BioRxiv [preprint]).

neuroscience↗
bioRxiv2026

The Genesis Framework: A Systems-Control Architecture for Safe Partial Cellular Rejuvenation via Stoichiometric saRNA Delivery, Integrated miRNA Logic Gating, and SORT Lipid Nanoparticles

BackgroundThe twelve hallmarks of aging as defined by Lopez-Otin et al. (Cell, 2023) represent the primary, antagonistic, and integrative mechanisms driving biological aging. To our knowledge, no open-source preclinical framework has yet attempted to coordinate all 12 official hallmarks in a single modular, non-integrating, safety-gated design validated across tens of thousands of computational cycles. This manuscript presents The Cepeda Framework -- a modular, safety-first preclinical research program designed to test whether coordinated partial rejuvenation strategies can be pursued in a falsifiable and systematically safety-constrained manner. Computational DevelopmentThe framework was developed through 21,000 dual-validation simulation cycles -- each simultaneously testing Efficacy (epigenetic reversal) and Safety (NANOG induction prevention) -- executed by the Forward Thinking Communities Scientific Group over one year of intensive computational modeling. The program evolved through four distinct phases: from the original Genesis Framework single-hallmark concept (Cycles 1-5,000; FAILURE -- 1:1:1 ratio unsafe), through stoichiometric ratio optimization (Cycles 5,001-12,000; PIVOT -- 3:2:1 ratio locked), gate verification (Cycles 12,001-18,000; SUCCESS -- dual-miRNA logic validated), and final 12-hallmark integration (Cycles 18,001-21,000; CONVERGENCE -- Cepeda Framework finalized). FrameworkThe Cepeda Framework integrates five regulatory principles -- (I) stoichiometric 3:2:1 OCT4:SOX2:KLF4 polycistronic m1{Psi}-saRNA; (II) 72-hour intrinsic saRNA pulse; (III) LEAD-7294 dual-miRNA logic gate (kill-switch + Let-7/L7Ae/Kt activation gate); (IV) NRF2/NAD+/paracrine microenvironmental co-treatment; (V) Genesis-TERT-01 transient hTERT telomere restoration -- with five companion protocols (Rapamycin pulse, Metformin, AP39 H2S donor [HIGH-UNCERTAINTY; Tier 1A required], GGA/4-PBA proteostasis stack, and Akkermansia/synbiotic/butyrate dysbiosis correction). Together, these components map to all 12 official Lopez-Otin 2023 hallmarks plus one Copenhagen 2022 candidate hallmark (RNA splicing dysregulation). SafetyNine parallel and sequential safety safeguards ensure no single-point failure produces uncontrolled reprogramming or sustained exposure, including the embedded LEAD-7294 miR-294 kill-switch (predicted t[1/2] <12 min at >5,000 copies/cell), the proactive Let-7/L7Ae/Kt activation gate, intrinsic saRNA transience, the 3:2:1 stoichiometric brake, and a predefined Acute Cytokine Response (ACR) monitoring protocol. All companion agents are clinically approved or have established in vitro/in vivo safety profiles. StatusTheoretical preclinical framework. No wet-laboratory validation of the integrated protocol has been conducted. All quantitative predictions require experimental falsification. Validation proceeds through a five-tier modular preclinical roadmap beginning with aged human dermal fibroblasts. All simulation code and archives will be released under CC0 at or before preprint submission.

molecular biology↗
bioRxiv2026

Evolutionary transfer learning enables organism-wide inference of mammalian enhancer landscapes

Understanding and modeling how a single human genome concurrently encodes gene regulatory programs for thousands of cell types remains a central challenge in genomics and machine learning. Most human cell types emerge during embryonic, fetal, and pediatric development which are inaccessible to comprehensive molecular profiling. To circumvent this, we hypothesized that the mismatch in evolutionary rates between cis-acting enhancers that modulate gene expression (fast) and the trans-acting regulatory factors that specify cell types (slow) creates an opportunity for evolutionary transfer learning. Specifically, models trained to predict cell type-specific enhancers in one species should generalize to the orthologous cell types and enhancers of related species. To test this, we generated a single-cell atlas of chromatin accessibility spanning mouse embryonic day 10 (E10) to birth (P0). Using combinatorial indexing1, we profiled 3.9 million nuclei from 36 staged embryos, resolving genome-wide accessibility in 36 cell classes and 140 cell types. We then trained a series of multi-output deep learning models (CREsted2), each addressing limitations of the preceding approach, towards the goal of genome-wide prediction of distal enhancers across major developmental lineages. An evolution-naive model achieved strong performance on heldout peaks, but exhibited two failure modes during genome-wide inference: overprediction at tandem repeats and conflation of promoter and distal enhancer grammars. An evolution-aware model resolved these by regrouping accessible regions based on their retention and functional coherence across mammalian evolution, but failed to generalize across species. Finally, an evolution-augmented model, STEAM (Synteny-aware Transfer learning for Enhancer Activity Modeling), incorporated enhancer orthologs from 241 mammalian genomes (Zoonomia3) in a synteny-supervised manner. This increased the effective data scale by as much as 195-fold, markedly improving generalization across mammals despite greater label noise. We applied STEAM to the genome-wide inference of cell class-specific distal developmental enhancers in humans, mice (HumMus) and 239 additional mammals3 (BabaGanoush), i.e. 32 x 241 = 7,712 genome-wide distal enhancer prediction tracks. Together, our results unify advances in single-cell profiling, deep learning, and comparative genomics into a framework for the evolutionary transfer learning of noncoding regulatory grammars. More broadly, our work supports the view that model organisms and evolutionarily diverse genomes are indispensable resources for accelerating and enhancing the AI-enabled exploration of human biology. NoteAn interactive version of this preprint, together with count matrices, CREsted models, prediction tracks, code and reproducible figures, is available at this link (ref 4).

genomics↗