Biology subjects

Gong, J.

Publications and source records attributed to Gong, J..

At least 19 recordsLinked to original sources

bioRxiv2026

Pretraining Enhances Megabase-Scale Gene Expression Prediction with GeneUnet

Predicting gene expression from DNA sequence across diverse genomic tracks is essential for understanding gene regulation and interpreting non-coding variants. Existing supervised methods are limited to few species and fail to exploit conserved regulatory mechanisms, while DNA foundation models capture cross-species information but remain constrained to kilobase-scale contexts insufficient for this task. Here we introduce GB.GeneUnet, an 837M-parameter transformer-based U-Net pretrained on 6 trillion tokens from multi-species genomes in OpenGenome2, extending genomic context to 1 Mb with up to 100x inference speedup over GeneMoE, a preliminary MoE transformer baseline of similar model size pretrained on the same data. Fine-tuned for gene expression prediction, GB.GeneUnet achieves state-of-the-art performance on the Borzoi benchmark at 524 kb context, and attains performance comparable to AlphaGenome at 1 Mb context while requiring a lighter fine-tuning procedure. Together, these results establish a scalable framework linking multi-species pretraining to ultra-long-context gene expression modeling.

genomics↗

bioRxiv2026

Time-Resolved Single-Cell Atlas Reveals Early Endothelial Activation and Stage-Dependent Immune-Stromal Communication in HFpEF

BackgroundHeart failure with preserved ejection fraction (HFpEF) is a heterogeneous syndrome associated with metabolic stress, hypertension, systemic inflammation, and microvascular dysfunction. Early cell-type-specific events and intercellular communication programs that accompany disease onset and progression remain poorly defined. MethodsWe performed a longitudinal study of HFpEF progression in high-fat diet (HFD)+L-NAME mice at control/baseline (0 weeks, 0w/Ctrl), early (1w), intermediate (4w), and established (8w) stages. Metabolic, hemodynamic, exercise, echocardiographic, and single-cardiomyocyte function were assessed. Cardiac non-cardiomyocytes (non-CMs) were profiled by single-cell RNA sequencing (scRNA-seq), with bulk RNA-seq for tissue-level comparison. Endothelial remodeling was assessed in an L-NAME-independent HFD plus mild transverse aortic constriction model (HFD+mTAC) and a published human HFpEF single-nucleus RNA-seq cohort. An endothelial-macrophage adhesion assay tested whether HFpEF-mimic stress promotes endothelial activation and macrophage adhesion. ResultsIn the HFD+L-NAME model, metabolic dysfunction, hypertension, reduced exercise tolerance, abnormal diastolic filling with preserved ejection fraction, and altered cardiomyocyte calcium handling were detected by 1w and persisted through 8w. Bulk RNA-seq showed progressive remodeling, with limited change between 8w and 12w, guiding scRNA-seq timepoint selection. scRNA-seq of 94,848 cardiac non-CMs identified nine major populations with stage-dependent remodeling. Endothelial cells (ECs) were recovered in high proportion and showed an early, pronounced transcriptional response, with inflammatory, adhesion, interferon-response, migratory, and vascular-remodeling programs emerging by 1w. Related EC activation signatures were observed in HFD+mTAC and human HFpEF data. Functionally, HFpEF-mimic stress increased adhesion and chemokine expression in human ECs and enhanced macrophage adhesion. Fibroblast matrix remodeling occurred at later stages, while macrophages progressively shifted toward inflammatory states. CellChat suggested stage-dependent communication remodeling from early endothelial-immune interactions toward later macrophage-fibroblast crosstalk. ConclusionTime-resolved scRNA-seq reveals coordinated, stage-dependent remodeling of the cardiac microvascular and interstitial microenvironment during HFpEF progression. Early endothelial activation emerges before later fibroblast matrix remodeling and inflammatory macrophage remodeling, identifying candidate cell states and signaling pathways for future mechanistic investigation. Clinical PerspectiveO_ST_ABSWhat Is New?C_ST_ABSO_LIThis study provides a time-resolved single-cell atlas of the cardiac non-cardiomyocyte compartment across baseline, early, intermediate, and established stages of HFpEF progression, rather than a single late-stage snapshot. C_LIO_LIEndothelial cells exhibit early inflammatory, adhesion, interferon-response, and vascular-remodeling programs within the first week of disease, preceding the later predominance of fibroblast matrix remodeling and inflammatory macrophage remodeling. C_LIO_LIThis endothelial activation signature is supported across two mechanistically distinct HFpEF mouse models and aligns with endothelial inflammatory and vascular-remodeling programs in human HFpEF myocardium, supporting its translational relevance. C_LI What Are the Clinical Implications?O_LIEarly endothelial activation may represent a targetable stage of HFpEF pathogenesis that arises before more established structural and fibrotic remodeling. C_LIO_LITherapeutic strategies aimed at limiting endothelial inflammatory activation or endothelial-immune interactions may help attenuate downstream vascular, immune, and stromal remodeling in HFpEF. C_LIO_LIThese findings provide a preclinical foundation for future longitudinal human studies testing whether early endothelial activation can serve as a biomarker, therapeutic target, or disease-staging feature in HFpEF. C_LI

cell biology↗

bioRxiv2026

A systematic imputation framework for sparse, multimodal space biology datasets: application to retinal imaging and omics from the RR9 mission

Missing data is a fundamental challenge in space biology, where high experimental costs, limited sample availability, and tissue allocation constraints produce datasets that are sparse, multimodal, and heterogeneous. We present a systematic four-stage framework for diagnosing, implementing, and validating data imputation strategies tailored to these characteristics, and demonstrate its application to retinal imaging and omics data from the NASA Rodent Research 9 (RR9) mission. Using logistic regression-based missingness diagnosis, we identify a Missing At Random (MAR) mechanism driven by experimental design constraints across nine assay modalities. We implement and optimize three imputation strategies: K-Nearest Neighbors (KNN), Multiple Imputation by Chained Equations with weak ElasticNet regularization (MICE-Elastic), and a per-column hybrid strategy, evaluated against a random sample imputer baseline. Validation across seven complementary metrics including supervised classification, unsupervised clustering, correlation structure preservation, masked value recovery, cross-dataset generalization, and permutation testing reveals that MICE-Elastic and the Hybrid strategy preserve genuine biological signal in both RNA-seq and TUNEL modalities, while KNN and the random sample imputer do not despite achieving comparable cross-validation accuracy. A critical finding is that imputation substantially improves supervised classification performance while consistently degrading unsupervised clustering structure, a trade-off researchers must understand before applying these methods. This framework provides practical, actionable guidance for space biologists and data scientists managing sparse multimodal datasets, and represents a foundational step toward digital twin development for space medicine.

bioinformatics↗

bioRxiv2026

Cross-Species Adaptation of RETFound for Rodent OCT Age Estimation Reveals Strong CNN Baselines in Data-Scarce Space Biology

Space-biology imaging studies are often constrained by severe data scarcity, limiting the development of robust machine-learning biomarkers. Rodent spaceflight and space-analog datasets provide an important preclinical setting for testing transfer-learning strategies, but the extent to which human retinal foundation models can generalize to rodent optical coherence tomography (OCT) remains unclear. Here, we benchmark cross-species adaptation of RETFound, a human retinal Vision Transformer pretrained on 1.6 million retinal images, for chronological age prediction from Brown Norway rat OCT B-scans in the NASA Open Science Data Repository dataset OSD-679. We adapted RETFound using Low-Rank Adaptation (LoRA) and evaluated performance on control animals under matched 3-fold rat-level cross-validation. We compared RETFound+LoRA with a strong ImageNet-pretrained Xception baseline under matched protocols and included a scratch/random ViT as a supplementary negative-control architecture check. Metrics included mean absolute error (MAE), R2, and inter-eye mean absolute difference (MAD). RETFound+LoRA achieved MAE = 26.20 {+/-} 5.03 days with R2 = 0.744 {+/-} 0.049. However, Xception performed better in the primary benchmark (MAE = 19.01 {+/-} 7.67 days, R2 = 0.853 {+/-} 0.082), and the matched-fold comparison favored Xception, although this result should be interpreted cautiously given the small number of folds. Inter-eye consistency was maintained across the matched control evaluation, and saliency maps localized model attention to anatomically plausible inner retinal regions. Together, these results show that human retinal foundation models can transfer to rodent OCT in a scientifically useful way, but also that strong CNN baselines may outperform transformer-based models in small-sample cross-species settings. This preprint provides a reproducible benchmark and baseline framework for future retinal biomarker development in space biology. Significance StatementSpace-biology imaging studies are intrinsically data-limited. This preprint provides a reproducible cross-species benchmark for adapting Earth-trained retinal models to rodent OCT under small-sample conditions, highlights the value of strong CNN baselines, and offers a reusable starting point for future retinal biomarker development in space-relevant datasets.

bioinformatics↗

bioRxiv2026

Harnessing AI to Build Virtual Cells

A virtual cell is a world model of a cell: a computational system that predicts, simulates and programs cellular processes across modalities and scales. An important path toward this goal is to model how genetic and chemical perturbations give rise to transcriptional responses, a core capability for disease understanding and drug discovery. However, current approaches remain expert-intensive, relying on iterative manual model design, training and debugging over months. Here we present VCHarness, an autonomous AI system that constructs perturbation-response models by combining an AI coding agent with multimodal biological foundation models. The system explores large spaces of architectures and training pipelines with minimal human intervention, iteratively generating, evaluating and refining candidate models. Across multiple perturbation-response benchmarks, VCHarness identifies architectures that outperform expert-designed approaches while reducing development time from months to days. It further uncovers non-obvious architectural patterns associated with improved performance, indicating that automated search can extend beyond conventional design strategies. These results suggest a shift from manually engineered models toward autonomous systems for constructing components of virtual cell world models, enabling scalable and data-driven exploration of cellular systems.

bioinformatics↗

bioRxiv2025

Hypobaric hypoxia drives citrate cycle reprogramming to suppress tumor progression

High altitude residents present lower tumor incidence and mortality than plain residents, suggesting that hypobaric hypoxia displays protective effect against tumor development. The results of our studies showed that hypobaric hypoxia exposure significantly suppressed tumor progression in subcutaneous bearing and lung metastasis tumor models. With single-cell transcriptome analysis, we identified HIF-1 pathway was significantly downregulated in tumor tissues with high altitude hypoxia, leading to decreased glycolysis and angiogenesis and improved antitumor immune response. Mechanistically, systemic hypoxia rewired citric acid cycle, with enhanced CS and IDH2 expression and reduced OGDH and SUCLG2 expression, to induce -ketoglutarate accumulation and succinate decline in tumor microenvironment, further mediating HIF-1 pathway inhibition. Moreover, hypobaric hypoxia treatment significantly improved the antitumor effects of adjuvant therapies (including chemotherapy, anti-angiogenic therapy and immunotherapy). Our findings reveal the role and mechanism of hypobaric hypoxia for tumor regression and yield new insights into tumor therapy.

cancer biology↗

bioRxiv2025

AIDO.Tissue: Spatial Cell-Guided Pretraining for Scalable Spatial Transcriptomics Foundation Model

Single-cell spatial transcriptomics enables high-resolution insights into tissue organization and cell-cell interactions, yet poses significant computational and modeling challenges due to its scale and complexity. Here we introduce AIDO.Tissue, a spatially-informed pretraining framework. The design employs multiple cells as input and an asymmetric encoder-decoder architecture, making it effectively encodes cross-cell dependencies while scaling to large data. Systematic evaluation shows that our method scales with neighboring size and achieves state-of-the-art performance across diverse downstream tasks, including spatial cell type classification, cell niche type prediction and cell density estimation. These results highlight the importance of spatial context in building general-purpose foundation models for tissue-level understanding.

bioinformatics↗

bioRxiv2025

Regulation of alternative polyadenylation events by PABPC1 affects erythroid progenitor cell expansion

Erythropoiesis is precisely regulated by multilayer networks and is crucial for maintaining steady-state hemoglobin levels and ensuring effective oxygen transport. Alternative polyadenylation (APA) is a post-transcriptional regulatory mechanism that generates multiple mRNA isoforms from a single gene, based on specific 3-untranslated region sequences. While APA plays a vital role in various cellular processes, its mechanism in erythropoiesis remains unexplored. Here, we employed an integrative approach combining bioinformatics and experimental validation to systematically analyze APAs role in erythropoiesis. We mapped the APA landscape during erythroid differentiation, discovering notable APA shifts crucial for the maturation from burst-forming unit erythroid to colony-forming unit erythroid. Notably, our findings highlighted PABPC1 as the primary APA regulator of these stages. Functional investigations revealed that knocking down PABPC1 disrupts erythroid progenitor cell proliferation and differentiation, implicating the proteins essential role in modulating cell fate through APA regulation. We further identified that decreased PABPC1 levels led to the increased usage of the proximal polyA site of TSC22D1 and gene overexpression, revealing a novel mechanism through which APA affects erythroid progenitor expansion and differentiation. These insights uncover a novel dimension of APA regulation in early erythropoiesis, providing new strategies for the treatment of diseases associated with erythropoiesis disorders.

developmental biology↗

bioRxiv2024

Balancing Locality and Reconstruction in Protein Structure Tokenizer

The structure of a protein is crucial to its biological function. With the expansion of available protein structures, such as those in the AlphaFold Protein Structure Database (AFDB), there is an increasing need for efficient methods to index, search, and generate these structures. Additionally, there is a growing interest in integrating structural information with models from other modalities, such as protein sequence language models. We present a novel VQ-VAE-based protein structure tokenizer, AIDO.StructureTokenizer (AIDO.St), which is a pretrained module for protein structures in an AI-driven Digital Organism [1]. AIDO.StructureTokenizer is a 300M parameter model consisting of an equivariant encoder to discretize input structures into tokens, and an invariant decoder to reconstruct the inputs from these tokens. In addition to evaluating structure reconstruction ability, we also compared our model to Foldseek, ProToken, and ESM3 in terms of protein structure retrieval ability. Through our experiments, we discovered an intriguing trade-off between the encoders locality and retrieval ability and the decoders reconstruction ability. Our results also demonstrate that a better balance between retrieval and reconstruction enables a better alignment between the structure tokens and a protein sequence language model, resulting in better structure prediction accuracy. Models and code are available through ModelGenerator in https://github.com/genbio-ai/AIDO and on Hugging Face.

bioinformatics↗

bioRxiv2024

CellBinDB: A Large-Scale Multimodal Annotated Dataset for Cell Segmentation with Benchmarking of Universal Models

In recent years, cell segmentation techniques have played a critical role in the analysis of biological images, especially for quantitative studies. Deep learning-based cell segmentation models have demonstrated remarkable performance in segmenting cell and nucleus boundaries, however, they are typically tailored to specific modalities or require manual tuning of hyperparameters, limiting their generalizability to unseen data. Comprehensive datasets that support both the training of universal models and the evaluation of various segmentation techniques are essential for overcoming these limitations and promoting the development of more versatile cell segmentation solutions. Here, we present CellBinDB, a large-scale multimodal annotated dataset established for these purposes. CellBinDB contains more than 1,000 annotated images, each labeled to identify the boundaries of cells or nuclei, including 4,6-Diamidino-2-Phenylindole (DAPI), Single-stranded DNA (ssDNA), Hematoxylin and Eosin (H&E), and Multiplex Immunofluorescence (mIF) staining, covering over 30 normal and diseased tissue types from human and mouse samples. Based on CellBinDB, we benchmarked seven state-of-the-art and widely used cell segmentation technologies/methods, and further analyzed the effects of four cell morphology indicators and image gradient on the segmentation results.

cell biology↗

bioRxiv2024

Multi-year drought strengthens positive and negative functional diversity effects on tree growth response

O_LIMixed-species forests are proposed as strategy to increase the resistance and resilience of forests to drought stress. However, evidence suggest that increasing tree species richness does not consistently enhance tree growth responses to drought. Moreover, tree diversity effects under unprecedented multiyear droughts remain uncertain, calling for a better understanding of the underlying processes. C_LIO_LIHere, we used a network of planted tree diversity experiments to investigate how drought-induced growth responses of individual trees are influenced by neighborhood tree diversity and the functional traits of the focal tree species. We analyzed tree cores (948 trees across 16 species) from nine experiments across Europe featuring gradients of tree species richness (1-6 species), which experienced severe droughts in recent years. Radial growth response to drought was quantified as tree-ring biomass increment using X-ray computed tomography. We applied hydraulic trait-based growth models to analyze single-year drought responses across all sites and site-specific responses during consecutive drought years for six sites as a function of neighborhood tree diversity. C_LIO_LIThe large variability in tree growth responses to a single-year drought was partially explained by the focal species hydraulic safety margin (representing species drought tolerance) and drought intensity, but independent of neighborhood species richness or functional trait diversity. However, tree diversity effects on growth responses strengthened during consecutive drought years and were site-specific, with contrasting direction (both positive and negative). This indicated opposing pathways of diversity effects under consecutive drought events, possibly resulting from competitive release or greater water consumption in diverse mixtures. C_LIO_LIWe conclude that tree diversity effects on growth responses to single-year droughts may differ considerably from responses to consecutive drought years. Our study highlights the need to consider trait-based approaches (specifically, hydraulic traits) and tree neighborhood scale processes to understand the multifaceted growth responses of tree mixtures under prolonged drought stress. C_LI

ecology↗

bioRxiv2024

Targeting SLC7A11-mediated cysteine metabolism for the treatment of trastuzumab resistant HER2 positive breast cancer

Trastuzumab resistance remains a challenge for HER2 positive breast cancer treatment. Targeting metabolic reprogramming would provide novel insights for therapeutic strategies. Here, we integrated metabolomics, transcriptomics and epigenomics data of trastuzumab sensitive and primary resistant HER2 positive breast cancer to identify metabolic alterations. Aberrant cysteine metabolism was discovered in trastuzumab primary resistant breast cancer at both circulating and intracellular levels. The inhibition of SLC7A11 and cysteine starvation could synergize with trastuzumab to induce ferroptosis. Mechanistically, increased H3K4me3 and decreased DNA methylation enhanced SLC7A11 transcription and cystine uptake in trastuzumab resistant breast cancer. The regulation of epigenetic modifications modulated cysteine metabolism and ferroptosis sensitivity. These results revealed an innovative approach for overcoming trastuzumab resistance by targeting specific amino acid metabolism.

cancer biology↗

bioRxiv2024

Loss of SARM1 protects against retinal ganglion cell degeneration in Autosomal Dominant Optic Atrophy

Autosomal Dominant Optic Atrophy (ADOA), the most prevalent inherited optic neuropathy, leads to retinal ganglion cell (RGC) degeneration and vision loss. ADOA is primarily caused by mutations in the OPA1 gene, which encodes a conserved GTPase important for mitochondrial inner membrane dynamics. To date, the disease mechanism remains unclear, and no therapies are available. Here, we present a novel mouse model carrying the pathogenic Opa1R290Q/+ allele that recapitulates key features of human ADOA, including mitochondrial defects, age-related RGC loss, optic nerve degeneration, and reduced RGC functions. We identify SARM1, a neurodegeneration switch, as a key driver of RGC degeneration in these mice. Sarm1 knockout nearly completely suppresses all the degeneration phenotypes. Additionally, we show that SARM1 is located within the mitochondrial intermembrane space (IMS). These findings indicate that SARM1 is activated downstream of mitochondrial dysfunction in ADOA, highlighting it as a promising therapeutic target.

neuroscience↗

bioRxiv2024

A male seminal fluid protein SFP-1 regulates hermaphrodite post-mating longevity and fat metabolism in Caenorhabditis elegans

There are several causes of mating-induced physiological changes in hermaphrodites exposed to males, such as functional male sperms causing shrinking and the male pheromone mediating shortened lifespan, which utilize different molecular pathways and are shared across species. However, it is unclear whether the male seminal fluid protein contributes to this post-mating regulation. Here, we investigated the transmit way and the impacts of the Caenorhabditis male seminal fluid protein SFP-1 in mated hermaphrodite tissues. We find that SFP-1 is the key component in seminal fluid to induce post-mating physiological changes in mated hermaphrodites. It acts as a cargo packing into exophers which require the phospholipid scramblase ANOH-1 and ANOH-2 to develop in male seminal vesicle. Exophers carrying SFP-1 cross over the somatic gonad uterus and eventually the protein is uptaken by the intestinal cells via endocytosis. Within the intestine, the NTF2-like domain of SFP-1 assists the association and interaction with the transcription factors SKN-1 and DAF-16 to induce post-mating somatic fat depletion and a shortened lifespan. Our study reveals the elaborate strategies of the male seminal fluid protein on triggering mating-induced physiological changes elicited by sexual interactions that could exist in other species.

molecular biology↗

bioRxiv2024

Differential regulation of BAX and BAK apoptotic activity revealed by a novel small molecule

Defective apoptosis mediated by BAK or BAX underlies various human pathologies including autoimmune and degenerative conditions. The mitochondrial channel protein VDAC2 interacts with BAK and BAX through a common interface to either inhibit BAK or to facilitate BAX apoptotic activity. Using a newly developed small molecule (WEHI-3773) that inhibits the interaction between VDAC2 and BAK or BAX, we reveal contrasting effects on BAX and BAK apoptotic activity. WEHI-3773 inhibits apoptosis mediated by BAX by blocking VDAC2-mediated BAX recruitment to mitochondria. Conversely, WEHI-3773 primes BAK for apoptosis by impairing its inhibitory sequestration by VDAC2 on the mitochondrial membrane. In cells expressing both BAX and BAK, repressing their association with VDAC2 promotes apoptosis, because once BAK is activated, it further activates BAX through a feed-forward mechanism. In some leukemias, mutation or loss of BAX is a key driver of resistance to the BH3-mimetic anti-cancer drug venetoclax. Strikingly, promoting BAK-mediated killing by small molecule dissociation of the VDAC2 interaction can overcome this resistance in different leukemia models. These data reveal a hitherto unappreciated level of coordination of BAX and BAK apoptotic activity through their interaction with VDAC2 that may be targeted therapeutically.

cell biology↗

bioRxiv2024

CIB2 as a Ca2+ Sensor for Auditory Mechano-Electrical Transduction and Linked to Genetic-Heterogeneity of TMC1 in Hearing Loss

Non-syndromic sensorineural hearing loss is characterized by genetic heterogeneity, leading to potential clinical misdiagnosis. TMC1, a unique causative gene associated with deafness, exhibits variants with autosomal dominant and recessive inheritance patterns. TMC1 codes for the transmembrane channel-like 1 (TMC1), a key component of the mechano-electrical transduction (MET) machinery for hearing. However, the molecular mechanism of Ca2+ regulation in MET, which is essential for sound perception, remains unclear. CIB2, another MET component associated with deafness, can bind with calcium and may play a role in MET gating. Our study reveals that the TMC1-CIB2 complex undergoes a Ca2+-induced conformational change, highlighting the crucial role of Ca2+ in MET. We identified a vertebrate-specific binding site on TMC1, named CR3, that interacts with apo CIB2, with the binding interface linked to hearing loss. Using an in-vivo animal model, we demonstrated that disruption of the calcium-sensing site of CIB2 perturbs the MET channel conductivity, confirming its role as a Ca2+ sensor for MET. Additionally, after systematically analyzing the hearing-loss variants, we observed that dominant mutations of TMC1 are often located near the ion pore or at the binding interfaces with the Ca2+ sensor CIB2. This provides a detailed mechanism of the genetic heterogeneity of TMC1 in hearing loss at an atomic level.

molecular biology↗

bioRxiv2024

Optimization of gene knockout approaches and practical solutions to sgRNA selection challenges in hPSCs with inducible Cas9 system

RationaleCRISPR/Cas9 has been extensively used to knock out genes, allowing the study of genetic loss-of-function in human pluripotent stem cells (hPSCs). However, the current use of the Cas9-sgRNA plasmid or iCas9 system for gene editing in hPSCs has resulted in limited and inconsistent editing efficiency, as well as labor-intensive work. Additionally, identifying single-guide RNAs (sgRNAs) with high cleavage efficiency and distinguishing them from ineffective ones, which efficiently induce frameshift INDELs (Indels and Deletions) but fail to eliminate target proteins expression, are major challenges in gene knockout experiments. MethodsThis study addresses above issues using an optimized doxycycline-induced spCas9-expressing hPSCs (hPSCs-iCas9) system. We initially developed this system by optimizing a number of parameters to maximize INDELs introducing efficiency in hPSCs-iCas9 cells. The INDELs determined by this system were then compared to predicted scores from three cleavage efficiency scoring algorithms to validate the algorithms accuracy and consistency. Furthermore, we conducted gene knockout using a set of sgRNAs targeting different exons of the ACE2 gene to achieve approximately 80% INDELs for each targeting locus. Western blotting was then performed to detect ACE2 protein expression levels, enabling the identification of potentially ineffective sgRNAs. ResultsSeveral critical factors, including cell tolerance to nucleofection stress, sgRNA stability, nucleofection frequency, and the cell-to-sgRNA ratio, were found to have significant impact on editing efficiency in hPSCs-iCas9. Fine-tuning these parameters markedly improved this efficiency, resulting in up to 93% INDELs for single gene knockout. The three scoring algorithms exhibited significant differences or even conflicts in scoring cleavage efficiency. Through comparing experimental observations to predicted scores, we discovered that the Benchling algorithm outperformed the other two in terms of accuracy and consistency. Furthermore, a sgRNA targeting exon 2 of ACE2 gene was quickly identified as ineffective, as evidenced by the edited cells pool containing 80% INDELs while ACE2 protein expression retained unchanged detected by Western blot. ConclusionThe findings of this study offer valuable insights into the optimal design of gene knockout experiments in hPSCs and provide practical solutions to sgRNA selection challenges for gene editing.

cell biology↗

bioRxiv2024

Simulating 500 million years of evolution with a language model

More than three billion years of evolution have produced an image of biology encoded into the space of natural proteins. Here we show that language models trained on tokens generated by evolution can act as evolutionary simulators to generate functional proteins that are far away from known proteins. We present ESM3, a frontier multimodal generative language model that reasons over the sequence, structure, and function of proteins. ESM3 can follow complex prompts combining its modalities and is highly responsive to biological alignment. We have prompted ESM3 to generate fluorescent proteins with a chain of thought. Among the generations that we synthesized, we found a bright fluorescent protein at far distance (58% identity) from known fluorescent proteins. Similarly distant natural fluorescent proteins are separated by over five hundred million years of evolution.

synthetic biology↗