Biology subjects

Church, G.

Publications and source records attributed to Church, G..

At least 19 recordsLinked to original sources

bioRxiv2026

Inference and validation of causal genetic interactions in aging human fibroblasts

In the past hundred years, the fraction of fatalities due to age-related disease has risen from an envied minority of cases to the leading cause of death and economic morbidity in the developed world. Despite notable advances in understanding aging, its intrinsic complexity has prevented major therapeutic breakthroughs. This complexity extends across all systems and scales in an organism, leaving a "fingerprint of aging" across every aspect of cell and tissue behavior whose interrelationships are difficult to untangle. Compounding this problem, human aging is observed largely through cross-sectional rather than longitudinal data. Whether such data retain sufficient dynamical information to reveal causal interactions is unclear. Here we show that they can. Using convergent cross mapping (CCM), a method for identifying causal interactions in nonlinear dynamical systems, we reconstruct directed genetic interactions from age-ordered, cross-sectional transcriptomic data - recovering dynamics from populations rather than individuals. Applied to dermal fibroblasts from 490 GTEx donors, CCM identifies a genome-scale causal network whose topology, together with age-dependent changes in gene expression, predicts candidate drivers of aging. We tested these predictions experimentally by single-gene overexpression in fibroblasts from 55-year-old donors. Single-cell RNA sequencing confirmed 53.8% of testable CCM-predicted causal interactions, showing that dynamics reconstructed from cross-sectional populations contain experimentally recoverable information about gene-gene interactions. This 53.8% additionally exceeded that produced by multiple comparanda methods and a random null. We then asked whether this information could identify interventions that reverse cellular aging. Independent transcriptomic and high-dimensional morphological assays show that CCM-selected perturbations shift old fibroblasts toward states characteristic of young cells. These two measures of rejuvenation are strongly correlated, and the strongest interventions cause a majority of 55-year-old cells to resemble cells from 25-year-old donors. Thus, causal dynamics can be recovered from cross-sectional population data and used to identify interventions in aging. More generally, the results suggest a way to distinguish genes that change with age from genes that drive those changes, providing experimentally testable targets for altering the aging phenotype.

systems biology↗

bioRxiv2026

A Vast Navigable Protein Sequence Subspace

Fitness landscapes govern evolution. Our empirical understanding of their structure has primarily been informed by evolutionary history. Here, we combine high-throughput DNA synthesis and machine learning, without training on evolutionary priors, to build a high-resolution map of the genome packaging landscape of the Adeno-associated Virus capsid protein. We assay 232,123 mutants around Adeno-Associated Virus 2 (AAV2) including 15 neighborhoods centered on seed sequences within 5-21 mutations of the wild type sequence. Local neighborhoods differ sharply in mutational tolerance and epistatic detail, yet much of the quantitative variation is captured by a simple pattern: sitewise effects transformed by global epistasis. Strikingly, using targeted sampling of only a few thousand mutants, we uncover a vast percolating network of function-preserving ridges embedded within a largely non-functional landscape. These ridges support billions of distinct trajectories of up to 667 steps that connect all sampled neighborhoods through functional single mutations. Designed variants with mutations distinct from the evolutionary background retain packaging at distances where randomly generated mutants rarely survive. These include capsids that combine up to 7 mutations that are largely absent from the evolutionary record, and which are each individually lethal in the 15 diverse backgrounds we tested. Our results reveal rich subspace structure resembling landscape orebodies: reservoirs of highly viable and navigable sequence space defined by simple constraints. These subspaces can be discovered through targeted sampling and can open access to vast troves of proteins that may be valuable for downstream applications such as therapeutics.

evolutionary biology↗

bioRxiv2026

PG-LLM: Benchmarking General-Purpose Language Models for Protein Variant Ranking

General-purpose frontier language models are being increasingly utilized for protein-design work, yet their ability to understand and evaluate variant effects remains unclear. Here, we introduce PG-LLM, a benchmark comprising 276 protein-variant prioritization tasks: 217 from ProteinGym and a temporally held-out set of 59 from recently published studies. Each task follows the same format: a language model is asked to rank a list of variant sequences given only the wild-type protein sequence and an assay description with no access to tools, multiple-sequence alignments, or protein structures. We evaluate thirteen language models and 95 published protein predictors on the same variants with the same evaluation metric. Claude Opus 5 (Max) and GPT 5.6 Sol (Max) are the best performing LLMs with Spearman correlations of{rho} = 0.406 and 0.402 respectively. Opus 5 outperforms 49 of 95 published protein predictors, including 41 of 46 sequence-only methods, and approaches ESM2-650M at{rho} = 0.411, but remains below the leading predictor VenusREM at{rho} = 0.523. We observe that variant-ranking performance scales with test-time compute across GPT, Claude, and Gemini models, but gains taper before closing the gap to specialist protein predictors. To address contamination risk, we create a held-out evaluation set with 59 DMS assays from 19 studies whose scores first became public after January 2026. On this set, we observe performance and test time compute scaling trends similar to those on the 217 tasks derived from ProteinGym. PG-LLM shows that tool-free language models capture substantial protein-variant signal, outperforming many sequence-based predictors while remaining below the strongest specialized models.

bioinformatics↗

bioRxiv2026

Designing protein/non-protein binding interactions using a full-atom diffusion model

An unresolved challenge in the field of computational protein design is to create proteins that bind non-protein partners, e.g. DNA, RNA, and small molecules. Most machine learning (ML) algorithms for protein design can only work with systems composed entirely of amino acids, and therefore cannot be directly applied to this task. The few algorithms that accommodate non-proteins still represent amino acids differently than other molecules, and therefore cannot easily recognize the similarity between a sidechain and a small molecule that share a functional group. We introduce a new method, called AtomPaint, that avoids these limitations by employing a fully-atomic representation of protein structure. Starting from a model of a desired binding interaction, our method proceeds by (i) converting that model to a 3D image, (ii) masking out the parts of that image that need to be redesigned, (iii) using a diffusion model to inpaint the masked voxels, then (iv) using a classification model to identify the amino acids in the inpainted image. Both models are SE(3)-equivariant ResNets, and were trained on a dataset of structures from the Protein Data Bank (PDB) curated to emphasize protein/non-protein interactions. In a sequence recovery benchmark, AtomPaint performed better than random guessing, suggesting that it understands some aspects of molecular structure. We discuss possible avenues of improvement, in the hopes that the advantages of our novel image-based approach can be fully realized.

bioinformatics↗

bioRxiv2025

Engineered translation release factor 1 suppresses disease-causing nonsense mutations and enables multi-site nonstandard amino acids incorporation

Nonsense mutations convert sense codons into premature termination codons (PTC), resulting in early termination of translation of mRNAs, underlie [~]11% of human genetic diseases. Restoring translation of genes carrying nonsense mutations remains a major therapeutic challenge. Here we present OPENER (Overexpression of Protein-engineered eRF1 for Nonsense Elision and Readthrough), a strategy to efficiently suppress nonsense mutations based on engineered variants of the eukaryotic translation termination factor eRF1. Through saturation mutagenesis of the human eRF1 N-domain, we identified variants, notably eRF1S77E, that promote efficient readthrough of all three PTCs (UAG, UGA, UAA) and rescue 12 pathogenic PTCs reporters in multiple disease contexts. Systemic delivery of OPENER-S77E via AAV8 in a Duchenne muscular dystrophy mouse model (DmdQ995X) restored dystrophin expression, improved muscle function, and ameliorated pathology without detected toxicity. Furthermore, when combined with orthogonal engineered aminoacyl-tRNA synthetase/tRNA pairs, OPENER-S77E enabled efficient, multi-site incorporation of nonstandard amino acids (BocK and AzF) into a single protein in mammalian cells, achieving efficiency of 27.77% and 19.83%, respectively. The incorporation efficiency of BocK was further enhanced to 32.80% in conditional eRF1-knockout cells. OPENER thus provides a platform for treating nonsense-mediated diseases and expands the synthetic biology toolkit for precise protein engineering.

synthetic biology↗

bioRxiv2025

XTree enables memory-efficient, accurate short and long sequence alignment to millions of genomes across the tree of life

XTree is a k-mer-based aligner enabling rapid, memory-efficient alignment of sequencing reads to whole-genome reference databases with up to millions of genomes. Here, we detail XTrees performance on short and long read sequencing data and demonstrate its high accuracy across diverse bacterial, viral, and eukaryotic genomes. Benchmarking demonstrates superior and/or comparable precision and recall over existing tools, with more scalable indexing and efficient memory mapping. We additionally provide pre-indexed databases, including (1) the Genome Taxonomy Database (versions r214-r226), (2) representative GenBank fungi and protozoan genomes and (3) the Pan-Viral-Compendium, a bespoke data resource spanning 6.6 million, quality-controlled, viral genomes.

bioinformatics↗

bioRxiv2025

Defining bottlenecks and physiological impact of an orthogonal translation initiation system

Reengineering translation initiation provides a powerful route to develop new translation systems that enable precise control of protein synthesis. While many engineered translation systems show promise, their orthogonality and impact on host physiology is largely uncharacterized, limiting broader application. Here, we develop an initiator-tRNA with an AAC anticodon mutation (i-tRNA-AAC), enabling translation initiation at a GUU start codon. Using fluorescence assays, proteomics, and tRNA sequencing, we assess the i-tRNA-AAC initiation orthogonality and effects on the host, guiding the optimization of its translational efficiency. We find the i-tRNA-AAC mutant initiates translation exclusively from its GUU start codon and is improved by overexpression of valyl-tRNA synthetase and methionyl-tRNA formyltransferase. However, this intervention perturbs aminoacylation and base modifications of endogenous tRNAs along with proteome-wide changes likely due to increased valine demand. Our findings demonstrate how an orthogonal translation initiation system reshapes host physiology and reveals the adaptive responses that accompany translational reprogramming.

synthetic biology↗

bioRxiv2025

The Spatial Atlas of Human Anatomy (SAHA): A Multimodal Subcellular-Resolution Reference Across Human Organs

The Spatial Atlas of Human Anatomy (SAHA) represents the first multimodal, subcellular- resolution reference of healthy adult human tissues across multiple organ systems. Integrating spatial transcriptomics, proteomics, and histological features across over 15 million cells from more than 100 donors, SAHA maps conserved and organ-specific cellular niches in gastrointestinal and immune tissues. High-resolution profiling using CosMx SMI, 10x Xenium, RNAscope, GeoMx DSP, and single-nucleus RNA-seq reveals spatially organized cell states, rare adaptive immune populations, and tissue-specific cell-cell interactions and ligand-receptor pairs. Comparative analyses with colorectal cancer and inflammatory bowel disease demonstrate the power of SAHA to detect disease-associated spatial disruptions, including crypt dedifferentiation, perineural invasion, and therapy-resistant immune remodeling. All data are openly accessible through a FAIR-compliant interactive portal to support exploration, benchmarking, and machine learning model training. Through SAHA, we provide a foundational framework for spatial diagnostics and next-generation precision medicine grounded in a comprehensive human tissue atlas, enabling the development of context-aware models that simulate tissue behavior, decode complex pathologies, and accelerate therapeutic innovation at unprecedented scale.

systems biology↗

bioRxiv2025

On the ancestry and evolution of the extinct dire wolf

Dire wolves (Aenocyon dirus) are extinct predators of Pleistocene North America. Although phenotypically similar to living wolves (Canis lupus), dire wolves have yet to be placed confidently in the canid family tree. We generated 3.4x and 12.8x paleogenomes from two well-preserved dire wolves dating to > 13,000 and > 72,000 years ago, and estimated consensus species trees for these and 10 canid species. Our results revealed that [~]2/3 of dire wolf ancestry is derived from a lineage sister to the clade comprising the gray wolf, coyote, and dhole, and the remaining [~]1/3 from a lineage near the base of Canini diversity. We identified 80 genes evolving under diversifying selection in dire wolves. Our results underscore the power of paleogenomes to resolve long-standing taxonomic questions and contribute to growing evidence of the role of post-speciation gene flow as an evolutionary force.

evolutionary biology↗

bioRxiv2025

Programmable Nucleic Acid Sensing in Human Cells Using Circularizable ssDNA

Programmable technologies that sense specific nucleic acid signatures in living cells and trigger cellular functions hold significant potential for biotechnology and medicine. Here, we developed SONAR (Sensing Of Nucleic acids using ASOs and Reverse-transcriptases), a platform that enables the detection of target DNA and RNA sequences and triggers controlled gene expression in human cells. SONAR operates through circularizable single-stranded DNA (ssDNA) sensors that, upon hybridization with complementary DNA or reverse-transcribed RNA, undergoes target-dependent ligation via endogenous ligases, subsequently driving expression of genetic payloads. For RNA sensing, we employed chemically modified antisense oligonucleotides (ASOs) to prime targeted reverse transcription, generating complementary DNA that promotes ssDNA circularization. We demonstrate SONARs ability to detect DNA, exogenous and endogenous RNA transcripts, coupled with a programmable expression of diverse protein payloads, including reporters, recombinases, and genome editors. This platform establishes a versatile framework for targeted nucleic acid detection and inducible gene expression, with broad applications in diagnostics, therapeutics, and synthetic biology.

bioengineering↗

bioRxiv2024

Multiomic foundation model predicts epigenetic regulation by zero-shot

Deciphering cell type-specific perturbation effects on genes and cellular states demands considerable experimental resources. To overcome this challenge, we propose Perturbation Transformer (PertFormer), a foundation model to uncover functional and regulatory mechanisms through interpretable in silico perturbation process. PertFormer comprises 3 billion parameters pretrained on both bulk and single-cell datasets covering 9 types of multiomics (55 billion bp bulk multiomics and the largest-to-date 1.5 billion paired multiomic samples from 1 million single cells), capturing regulation of 300 kb around every genic region. PertFormer enables zero-shot prediction of functional perturbations and cell fate dynamics, outperforming existing methods by 20.9%-480.2%. PertFormer identifies novel tumor treatment targets, validated experimentally. The generalization capabilities of PertFormer have the potential to accelerate the discovery of biological and clinical targets.

genomics↗

bioRxiv2024

Human skin rejuvenation via mRNA

Aging is characterized by a gradual decline in function, partly due to accumulated molecular damage. Human skin undergoes both chronological aging and environmental degradation, particularly UV-induced photoaging. Detrimental structural and physiological changes caused by aging include epidermal thinning due to stem cell depletion and dermal atrophy associated with decreased collagen production. Here, we present a comprehensive single-cell atlas of skin aging, analyzing samples from young, middle-aged, and elderly individuals, including both sun-exposed and sun-protected areas. This atlas reveals age-related changes in cellular composition and function across various skin cell types, including epidermal stem cells, fibroblasts, hair follicles, and endothelial cells. Using our atlas, we have uncovered basal stem cells as a highly variable population across aging, more so than other skin cell populations, such as fibroblasts. In basal stem cells, we identified ATF3 as a novel regulator of skin aging. ATF3 is a transcriptional factor for genes involved in the aging process, with its expression reduced by 20% during aging. Based on this discovery, we developed an innovative mRNA-based treatment to mitigate the effects of skin aging. After treatment with ATF3 mRNA, cell senescence decreased 25%, and we observed an over 20% increase in proliferation in treated basal stem cells. Importantly, we also found communication between keratinocytes and fibroblasts as a critical component of therapeutic interventions, with ATF3 mRNA rescue of basal cells significantly enhancing fibroblast collagen production by approximately 200%. Furthermore, we validated the efficacy of ATF3 mRNA treatment in ex vivo human skin and in vivo mouse models. In ex vivo human skin, microneedle-mediated delivery of ATF3 mRNA induced robust rejuvenation, expanding the basal stem-cell population and enhancing dermal ECM reconstruction. In a wound healing mouse model, ATF3 mRNA reduced scarring, demonstrating strong regenerative potential for reversing age-related skin decline. We conclude that ATF3 mRNA treatment effectively reverses the effects of skin aging by modulating specific cellular mechanisms, offering a novel, targeted approach to human skin rejuvenation.

synthetic biology↗

bioRxiv2024

Evidence for in vivo mRNA Transport Between Mammalian Cells

The prevalence and significance of intercellular mRNA transport remains unknown. Direct detection of mRNA transfer between cells within an organism is challenging due to technical limitations associated with transgene encoded molecular labels and cell sorting techniques. In this study we analyzed human-to-mouse xenograft single-cell RNA sequencing data to identify mouse transcripts in recovered human cells. The murine transcriptome analysis implicates macrophages as a frequent mRNA donor cell type. We then developed an in vitro system using mouse RAW264.7 macrophages and human HeLa and HEK293 cells and confirmed the transfer of Ftl1 mRNA from mouse to human cells. Overall, our study provides compelling in vivo evidence for prevalent intercellular mRNA transfer in human-to-mouse xenograft models.

cell biology↗

bioRxiv2024

Induction of Meiosis from Human Pluripotent Stem Cells

An in vitro model of human meiosis would accelerate research into this important reproductive process and development of therapies for infertility. We have developed a method to induce meiosis starting from male or female human pluripotent stem cells. We demonstrate that DNMT1 inhibition, retinoid signaling activation, and overexpression of regulatory factors (anti-apoptotic BCL2, and pro-meiotic HOXB5, BOLL, or MEIOC) rapidly activates meiosis, with leptonema beginning at 6 days, zygonema at 9 days, and pachynema at 12 days. Immunofluorescence microscopy shows key aspects of meiosis, including chromosome synapsis and sex body formation. The meiotic cells express genes similar to meiotic oogonia in vivo, including all synaptonemal complex components and machinery for meiotic recombination. These findings establish an accessible system for inducing human meiosis in vitro.

synthetic biology↗

bioRxiv2024

Machine Learning and Directed Evolution of Base Editing Enzymes

As we enter the era of CRISPR medicines, base editors (BEs) emerged as one of the most promising tools to treat genetic associated diseases. However, unintended bystander editing beyond the target nucleotide poses a challenge to their translation into effective therapies. While many efforts have been made in the design of a universal enzyme with minimal bystander editing, the context dependent activity represents a major challenge for base editing-based therapies. In this work, we designed a sequence-specific guide RNA library with 3-extensions and detected guides that were able to reduce bystander and increase editing efficiency in a context dependent manner. The best candidate was later used for phage assisted non-continuous evolution to find a new generation of precise base editors. Simultaneously, we use protein language models trained on massive protein sequence datasets to find the evolutionarily plausible mutational patterns that can improve deaminase activity and precision. Both strategies provide a collection of precise TadA variants that not only drastically reduced bystander edits, but also was not in detriment of on-target activity. Our findings introduce a guide/enzyme parallel engineering pipeline, which lays the foundation for the development of new personalized genome editing strategies, ultimately enhancing the safety and precision of this groundbreaking technology.

synthetic biology↗

bioRxiv2024

Post-translation digital data encoding into the genomes of mammalian cell populations

High resolution cellular signal encoding is critical for better understanding of complex biological phenomena. DNA-based biosignal encoders alter genomic or plasmid DNA in a signal dependent manner. Current approaches involve the signal of interest affecting a DNA edit by interacting with a signal specific promoter which then results in expression of the effector molecule (DNA altering enzyme). Here, we present the proof of concept of a biosignal encoding system where the enzyme terminal deoxynucleotidyl transferase (TdT) acts as the effector molecule upon directly interacting with the signal of interest. A template independent DNA polymerase (DNAp), TdT incorporates nucleotides at the 3 OH ends of DNA substrate in a signal dependent manner. By employing CRISPR-Cas9 to create double stranded breaks in genomic DNA, we make 3OH ends available to act as substrate for TdT. We show that this system can successfully resolve and encode different concentrations of various biosignals into the genomic DNA of HEK-293T cells. Finally, we develop a simple encoding scheme associated with the tested biosignals and encode the message "HELLO WORLD" into the genomic DNA of HEK-293T cells at a population level with 91% accuracy. This work demonstrates a simple and engineerable system that can reliably store local biosignal information into the genomes of mammalian cell populations.

synthetic biology↗

bioRxiv2024

DERIVATION OF ELEPHANT INDUCED PLURIPOTENT STEM CELLS

The crisis of biodiversity loss in the anthropogenic era requires new tools for studying non-model organisms. Elephants, for example, are both an endangered species and excellent models studying complex phenotypes like size, social behavior, and longevity, but they remain severely understudied. Here we report the first derivation of elephant (Elephas maximus) induced pluripotent stem cells (emiPSCs) achieved via a two-step process of chemical-media induction and colony selection, followed by overexpression of elephant transcription factors OCT4, SOX2, KLF4, MYC {+/-} NANOG and LIN28A, and modulation of the TP53 pathway. Since the seminal discovery of reprogramming by Shinya Yamanaka, iPSCs from many species including the functionally extinct northern white rhinocerous have been reported, but emiPSCs have remained elusive. While for multiple species the reprogramming protocol was adopted with little changes compared to model organisms like mouse and human, our emiPSC protocol requires a longer timeline and inhibition of TP53 expansion genes that are hypothesized to confer unique cancer resistance in elephants. iPSCs unlock tremendous potential to explore cell fate determination, cell and tissue development, cell therapies, drug screening, disease modeling, cancer development, gametogenesis and beyond to further our understanding of this iconic megafauna. This study opens new frontiers in advanced non-model organism cellular models for genetic rescue and conservation.

cell biology↗

bioRxiv2024

Genome streamlining to improve performance of a fast-growing cyanobacterium Synechococcus elongatus UTEX 2973

Cyanobacteria are photosynthetic organisms that have garnered significant recognition as potential hosts for sustainable bioproduction. However, their complex regulatory networks pose significant challenges to major metabolic engineering efforts, thereby limiting their feasibility as production hosts. Genome streamlining has been demonstrated to be a successful approach for improving productivity and fitness in heterotrophs but is yet to be explored to its full potential in phototrophs. Here we present the systematic reduction of the genome of the cyanobacterium exhibiting the fastest exponential growth, Synechococcus elongatus UTEX 2973. This work, the first of its kind in a photoautotroph, involved an iterative process using state-of-the-art genome-editing technology guided by experimental analysis and computational tools. CRISPR/Cas3 enabled large, progressive deletions of predicted dispensable regions and aided in the identification of essential genes. The large deletions were combined to obtain a strain with 55 kb genome reduction. The strains with streamlined genome showed improvement in growth (up to 23%) and productivity (by 22.7%) as compared to the WT. This streamlining strategy not only has the potential to develop cyanobacterial strains with improved growth and productivity traits but can also facilitate a better understanding of their genome to phenome relationships. ImportanceGenome streamlining is an evolutionary strategy used by natural living systems to dispense unnecessary genes from their genome as a mechanism to adapt and evolve. While this strategy has been successfully borrowed to develop synthetic heterotrophic microbial systems with desired phenotype, it has not been extensively explored in photoautotrophs. Genome streamlining strategy incorporates both computational predictions to identify the dispensable regions and experimental validation using genome editing tool and in this study we have employed a modified strategy with the goal to minimize the genome size to an extent that allows optimal cellular fitness under specified conditions. Our strategy has explored a novel genome-editing tool in photoautotrophs which, unlike other existing tools, enables large, spontaneous optimal deletions from the genome. Our findings demonstrate the effectiveness of this modified strategy in obtaining strains with streamlined genome, exhibiting improved fitness and productivity.

synthetic biology↗