Abstract Here we present DynaVelo, a generative neural ordinary differential equation model that learns the joint dynamics of gene expression and transcription factor (TF) motif activity from single-cell multiome data. DynaVelo learns a functional form of the vector field, such that the model can be perturbed in silico and not only used to draw trajectories. We show that DynaVelo recovers the bifurcating in vivo dynamics of wild-type murine germinal center (GC) B cells and reveals how these dynamics change under heterozygous loss of epigenetic regulators Arid1a and Ctcf . In silico perturbation yields dynamic and cell-state-specific gene regulatory networks (GRNs) that recover known TF-to-gene edges in the wild-type GC GRN and predicts those that are disrupted in mutants. Perturbing Arid1aHet GCs in silico identified Ctcf , Stat3 and Bcl6 as rescue targets for Arid1a loss. In vivo experiments in double-heterozygous mice confirmed rescue of the Arid1aHet dark-zone phenotype in all three cases, and multiome profiling of Arid1aHet ; CtcfHet GCs validated the predicted shift toward wild-type dynamics.
Transcriptomic measurement technologies provide powerful approaches to understand biology but are limited to providing either a high-resolution snapshot at a moment in time or a low-resolution history of transcription constrained to a few genes over a narrow time range. Here, we introduce the ‘recordanelle’, a single-component synthetic organelle capable of recording the transcriptome in human cells. Recordanelles are composed of a mitochondrial membrane anchor, a condensate assembler and a poly(A)-binding protein. We demonstrate that the recordanelle can be engineered to recruit engineered RNAs or a sampling of up to 90% of all endogenous mRNAs and protect them over several days. We further show that transient and dynamic transcriptional responses resulting from inducible promoters or interferon signaling are preserved throughout time within the recordanelle at the single-cell level. This work provides a nondisruptive and unbiased transcriptome-wide single-cell RNA recording technology to enable a deeper understanding of biology. Recordanelles continuously and unbiasedly record the transcriptome throughout time in living mammalian cells.
Tools for manipulating RNA that rely on nonhuman proteins, such as ISC-based and Cas-based scaffolds, can lead to unintended and little-understood effects on human cells and present challenges in therapeutic translation. Replacing them with human proteins engineered for expanded, programmable functions could bypass the limitations of these foreign proteins. Here we demonstrate that human Argonaute 2 (AGO2) can be repurposed as a modular RNA targeting domain. AGO2 fused to a C-to-U deaminase enables AGO-led targeted editing of RNA (ALTER). Using guide RNAs that remodel target RNA structure for selective editing and reduced nuclease activity, we show that ALTER achieves, on average, 1.9-fold higher C-to-U editing on mRNA targets in HEK293T cells compared with Cas13 editors CURE and RESCUE. Transcriptome-wide RNA-seq revealed similar or lower levels of off-target editing compared to the same Cas13-based editors. Machine learning models predict that far fewer immunogenic peptides will result from ALTER and we show that ALTER is amenable to a validated workflow for artificial intelligence-directed deimmunization. Human Argonaute is repurposed for RNA editing.
Each year, Nature Biotechnology highlights companies that received sizeable early-stage funding in the previous year. Bioptimus is developing multimodal and multiscale AI pathology models for biomarker and drug treatment prediction.
Structural variants (SVs) are a major source of genomic diversity, yet their discovery remains challenging due to repetitive genomic contexts, alignment ambiguity and the trade-off between sequencing cost and read length. Here we introduce HitSV, which implements repetitiveness- and signature-density-aware breakpoint recognition coupled with precise haplotype-resolved local assembly, thereby enabling base-resolution SV reconstruction and genotyping in various sequencing technologies. On the HG002 benchmark, HitSV achieved 15–72% (long read), 10–36% (short read) and 12% (hybrid sequencing), respectively, higher genotype F1 scores than state-of-the-art SV callers. Applied to the 1KGP phase 4 cohort, HitSV expanded the SV callset by 31.5%, substantially reshaping allele-frequency landscapes. In a large Chinese long-read cohort, HitSV uncovered tandem-repeat-mobile-element composite arrays as a prevalent, multi-allelic class of complex SVs, highlighting composite repeat architectures as an important and previously underrecognized contributor to human structural variation. HitSV models the interaction between genomic repetitiveness and signature intensity to improve structural variant detection.
