Homology search plays a fundamental role in computational biology, enabling the identification of evolutionary relationships and functional similarities among biological sequences. However, current homology search method...Homology search plays a fundamental role in computational biology, enabling the identification of evolutionary relationships and functional similarities among biological sequences. However, current homology search methods, including BLAST, Foldseek and MMseqs2, often struggle to efficiently and accurately process the vast scale of biological databases. Here we introduce ERAST (efficient retrieval-augmented search tool), a solution designed to handle approximately 1 billion biological sequences within the largest vector database to date. ERAST combines large language models and vector database technology to provide both efficient and precise searches for homologous biological sequences. It enhances search quality by integrating preretrieval, retrieval and postretrieval optimization stages, and supports both nucleotide and protein sequences. Through advanced indexing techniques, fine-grained segmentation and metadata integration, ERAST achieves better precision while operating approximately 50 times faster than Foldseek and 50,000 times faster than TM-align. This performance allows ERAST to conduct accurate searches against billions of biological sequences in mere milliseconds. The vector database integrated with ERAST can be accessed at https://ai4s.tencent.com/erast .
Dual-objective 4Pi single-molecule localization microscopy (4Pi-SMLM) offers isotropic nanoscale resolution; however, its broader adoption is limited by instrumental complexity and stringent alignment requirements. Here...Dual-objective 4Pi single-molecule localization microscopy (4Pi-SMLM) offers isotropic nanoscale resolution; however, its broader adoption is limited by instrumental complexity and stringent alignment requirements. Here we introduce mirror-enhanced 4Pi-SMLM (me4Pi-SMLM), a single-objective configuration that uses mirror-based retroreflection of the illumination beam to generate phase-tunable interference fringes. This design improves the axial resolution of astigmatism-based methods by approximately fivefold, delivering performance comparable to conventional 4Pi-SMLM while greatly reducing system complexity and maintenance. me4Pi-SMLM achieves near-isotropic localization precision of 2-3 nm in biological samples, enabling clear and unambiguous visualization of diverse ultrastructural features. Furthermore, it achieves sub-15 nm isotropic resolution in brain slices and facilitates high-fidelity two-colour imaging, nanoscale whole-cell reconstruction and live-cell imaging. me4Pi-SMLM can be seamlessly integrated into existing 3D-SMLM systems, enhancing performance with minimal cost and effort.
Isakova A, Liu DD, Cvijović I
… +8 more, Sinha R, Eastman AE, Saul S, Detweiler AM, Neff N, Einav S, Weissman IL, Quake SR
Nat Biotechnol
· 2026 Mar · PMID 41917462
·
Full text
Current single-cell RNA atlases largely capture polyadenylated transcripts while missing critical regulatory layers from noncoding RNA. To address this, we develop a generalizable framework that adapts total RNA profilin...Current single-cell RNA atlases largely capture polyadenylated transcripts while missing critical regulatory layers from noncoding RNA. To address this, we develop a generalizable framework that adapts total RNA profiling for use in standard droplet-based platforms and captures a broad complement of coding and noncoding RNAs using a unified pipeline. Applying this approach to the developing human brain, we generate a dataset mapping diverse RNA biotypes across all neuronal and non-neuronal lineages, revealing biotype-specific expression programs with cell-type and temporal specificity. Tracking microRNA dynamics in Cajal-Retzius neurons, transient and early-born neurons in the cortex, we show the enrichment and target anticorrelation of MIR137, associated with schizophrenia and intellectual disability, suggesting tight regulatory control. We apply TotalX to human peripheral blood mononuclear cells and identify transcriptional modules combining coding and noncoding RNAs and tRNA dynamics. In addition, we analyze dengue-infected hepatocytes and capture non-adenylated viral transcripts that distinguish infection states. This expanded coverage helps with understanding cellular identity and gene regulation at the atlas scale.
Precise modulation of gene expression through cis-regulatory editing holds promise for nontransgenic crop improvement. However, the sequence-to-function relationships that govern plant promoter activity remain poorly und...Precise modulation of gene expression through cis-regulatory editing holds promise for nontransgenic crop improvement. However, the sequence-to-function relationships that govern plant promoter activity remain poorly understood. Here we develop a massively parallel reporter assay in Sorghum bicolor to systematically measure the effects of >30,000 mutations spanning deletions, substitutions and motif insertions accessible through CRISPR editing across entire native promoters and 5' untranslated regions of 3 photosynthesis genes: PsbS, Raf1 and SBPase. We find that gene expression is most tunable within a ~500-bp core promoter region. The mutational effects are reproducible across biological replicates and predictive of protein output. Within these regions, we identify compact deletions and motif insertions that strongly increase protein production (>30-fold relative to wild type), outperforming transgenic enhancer elements. Mutation-effect relationships are gene specific, highlighting the need for tailored regulatory maps. Our results establish a high-throughput strategy for cis-regulatory fine-mapping that may enable crop improvements through minimal, precise and nontransgenic gene edits.
Nat Biotechnol
· 2026 Mar · PMID 41896477
·
Full text
Long-read metagenome assembly promises complete genomic recovery from microbiomes. However, the complexity of metagenomes poses challenges. Here we present myloasm, a metagenome assembler for modern long reads such as Pa...Long-read metagenome assembly promises complete genomic recovery from microbiomes. However, the complexity of metagenomes poses challenges. Here we present myloasm, a metagenome assembler for modern long reads such as PacBio HiFi and Oxford Nanopore Technologies (ONT) R10.4 long reads. Myloasm uses polymorphic k-mers to construct a high-resolution string graph and then leverages differential abundance for graph simplification. On real-world ONT metagenomes, myloasm assembled three times more complete circular contigs than the next-best assembler. Myloasm can make ONT and HiFi assemblies comparable. For example, on a jointly sequenced gut metagenome, myloasm with ONT assembled more complete circular genomes than any assembler with HiFi. Myloasm also recovers previously inaccessible within-species diversity. Here, we recovered six complete Prevotella copri single-contig genomes from a gut metagenome and eight complete TM7 (Saccharibacteria) contigs with >93% similarity from an oral metagenome. Overall, we show that myloasm outperforms existing long-read metagenome assemblers across a range of environments and modern sequencing technologies.
Antibody-drug conjugates enable highly specific delivery of potent cytotoxics to biomarker-expressing cells. In parallel, advances in DNA circuitry and DNA-protein conjugates have allowed programmable integration of mole...Antibody-drug conjugates enable highly specific delivery of potent cytotoxics to biomarker-expressing cells. In parallel, advances in DNA circuitry and DNA-protein conjugates have allowed programmable integration of molecular inputs and signal amplification via hybridization chain reactions (HCRs). Here we present a system using affibody-DNA and aptamer-DNA conjugates to execute a Boolean logic operation on cell-surface biomarkers, resulting in amplified payload delivery using an HCR of DNA-drug conjugates. Proximity-induced assembly of the biomarker binders generates the initiator that triggers an HCR. The resulting assembly undergoes endocytosis, enabling controlled payload release of drugs conjugated to the DNA with cathepsin-cleavable linkers. We show that DNA-drug conjugates achieve targeted delivery with >100-fold amplification relative to the input biomarkers using fluorescence quantifications. We also identify payloads that strongly influence delivery efficiency and demonstrate delivery of different drug combinations. Finally, we show that biomarker-triggered HCRs can recruit generic antibodies. This modular technology enables tailored combinations of biomarker inputs and drug outputs toward more precise and personalized treatment.
Generalist biological artificial intelligence (GBAI) represents a transformative approach to modeling the 'language of life'-the flow of information from DNA to cellular function. This Review synthesizes rapid advances i...Generalist biological artificial intelligence (GBAI) represents a transformative approach to modeling the 'language of life'-the flow of information from DNA to cellular function. This Review synthesizes rapid advances in biological AI to interpret and generate DNA, RNA, proteins and cellular systems. We chart a course toward comprehensive systems that can concurrently process and predict across these domains, performing several critical biological tasks simultaneously. Substantial opportunities lie in synergizing language and structural AI, leveraging specialized models and improving AI agents for autonomous discovery. After addressing challenges in data, biological complexity, scalability and experimental validation, GBAI has the potential to deepen our understanding of disease pathways and biomarkers, advance automated therapeutic design and evaluation, and integrate within virtual cells to meaningfully simulate biological activity.
Tissue engineering of the esophagus has been limited by stent dependance and poor muscle regeneration. Here we report an integrated strategy to engineer a 2.5-cm esophageal segment by microinjecting autologous pericyte-l...Tissue engineering of the esophagus has been limited by stent dependance and poor muscle regeneration. Here we report an integrated strategy to engineer a 2.5-cm esophageal segment by microinjecting autologous pericyte-like myogenic precursors and fibroblasts in a decellularized porcine scaffold to repair circumferential defects in 10-kg minipigs (n = 8), modeling pediatric use. Bioreactor maturation induced a proangiogenic phenotype, with in vivo support from biodegradable intraluminal stents and a vascularizing pleural wrap. This coordinated approach yielded safe and effective esophageal conduits; oral feeding supported normal growth, morbidity resembled that of clinical esophageal replacement and was endoscopically manageable, and 63% (5/8) survived to the 6-month endpoint. Comprehensive multimodal analyses demonstrated progressive recapitulation of native architecture, with increasing neuromuscular regeneration and vascularization, correlating with functional recovery, absence of symptomatic stricture and the presence of secondary peristalsis by 6 months. These results demonstrate that the combination of complementary regenerative, conditioning and surgical strategies enables a functionally integrated, contractile esophageal graft with ongoing structural maturation without immunosuppression.
Despite advances in mass spectrometry and emerging single-molecule approaches, sequencing peptides at the single-molecule level remains a central challenge in proteomics. Here we present a 'reverse translation' strategy...Despite advances in mass spectrometry and emerging single-molecule approaches, sequencing peptides at the single-molecule level remains a central challenge in proteomics. Here we present a 'reverse translation' strategy that enables single-molecule peptide sequencing with single-amino-acid resolution. In this approach, peptides undergo a modified Edman degradation that iteratively releases N-terminal amino acids tagged with peptide-specific DNA barcodes. Antibody-mediated proximity extension assays identify these barcoded amino acids and generate PCR-amplifiable DNA reporters that record the identity, position and originating peptide of each amino acid. The resulting DNA library is directly read by high-throughput sequencing, converting peptide sequences into digital DNA outputs. Using this approach, we demonstrate true single-molecule peptide sequencing, achieving full sequence coverage in millions of reads and accurate differentiation of both native and post-translationally modified peptides. These results establish a framework that redefines protein sequencing as a DNA sequencing problem and lays the foundation for high-throughput, de novo single-molecule protein sequencing.
The initial development of adenine base editors (ABEs), which facilitate A•T to G•C base pair changes in the genome, used directed evolution to install 14 mutations into the wild-type deaminase TadA, producing the first-...The initial development of adenine base editors (ABEs), which facilitate A•T to G•C base pair changes in the genome, used directed evolution to install 14 mutations into the wild-type deaminase TadA, producing the first-of-its-kind editor ABE7.10. Here we study the installed mutations' impacts on TadA fitness using comprehensive reversion analysis and apply our results to engineer more efficient, precise editors. By measuring activity in both human and Escherichia coli host systems, we categorize mutations as critical, dispensable or host dependent. We show that up to five mutations can be reverted back to wild type, generating minimally evolved ABEs (ME-ABEs). ME-ABEs show narrow editing windows (similar to that of ABE7.10) and enhanced on-target editing (matching activities of the high-activity editor variants ABE8e and ABE8.20 in most sequence contexts) and exhibit low levels of guide-RNA-dependent and guide-RNA-independent off-target activity. ME-ABEs efficiently target six sites of clinical interest that had previously proved challenging to edit with ABE7.10, ABE8e or ABE8.20.
Tumor immunotherapy is often compromised by an immunosuppressive tumor microenvironment (TME) characterized by abnormal vasculature and exhausted T cells. Here, given the role of nitric oxide (NO) in favorably remodeling...Tumor immunotherapy is often compromised by an immunosuppressive tumor microenvironment (TME) characterized by abnormal vasculature and exhausted T cells. Here, given the role of nitric oxide (NO) in favorably remodeling the TME, we engineered Escherichia coli Nissle 1917 (ECN) with a synthetic arginine-NO circuit (ECN-NO) that modifies the arginine synthesis pathway to constitutively synthesize arginine and enable sustained NO production. Specifically, deletion of the arginine repressor ArgR relieved feedback inhibition of arginine biosynthesis, whereas co-expression of argininosuccinate synthase and lyase (ArgG/ArgH), together with Bacillus subtilis nitric oxide synthase (BsNOS), enabled sustained NO production through enhanced arginine regeneration. Intratumoral colonization of ECN-NO significantly enhanced the antitumor efficacy of anti-programmed cell death ligand 1 (αPD-L1) immunotherapy, resulting in durable tumor regression across multiple solid tumor mouse models. Mechanistically, ECN-NO induced vascular normalization and dendritic cell recruitment, alleviated tumor immunosuppression and synergized with αPD-L1 to expand functional CD8 T cells, reverse T cell exhaustion and promote memory T cell formation, establishing antitumor immunity for at least 120 days.
Generative modeling offers a robust framework for designing functional DNA, RNA and protein sequences. However, physical synthesis of these sequences at scale is prohibitively expensive. We introduce a method to efficien...Generative modeling offers a robust framework for designing functional DNA, RNA and protein sequences. However, physical synthesis of these sequences at scale is prohibitively expensive. We introduce a method to efficiently synthesize DNA designs from generative models. The method integrates machine learning and wet lab procedures, implementing generative sampling algorithms physically through controlled stochastic chemical reactions using DNA oligosynthesis. We synthesize ~10 designs from a generative model of human antibodies, with realism and diversity comparable to state-of-the-art protein language models, and verify the designed DNA by sequencing. Translation and expression of the designed antibody scFvs in human cell lines, combined with high-throughput screening against multiplexed human leukocyte antigen (HLA)-presented intracellular proteins, yields potential therapeutic chimeric antigen receptors. We further synthesize ~10 DNA designs from models of Taq polymerase and the HLA-presented peptidome, confirming the method's generalizability.