Search bioRxiv⌕ Search
Home/Archive/bioRxiv

bioRxiv · 10.64898/2026.10.29.762960

Molecule-resolved inference reveals the spatial organization of transcriptional programs

Si, Y.·Lee, C.·Cheng, W.·Sweeney, M. D.·Yun, J. H.·Kim, M.·Jun, G.·Lee, J. H.·Kang, H. M.
View original source ↗Export metadata ↓

Abstract

Transcriptional programs encode cellular identity and physiological activity, linking gene regulation to cellular phenotypes. Although spatial transcriptomics measures individual RNA molecules across intact tissues, most analyses aggregate them into cells or pixels, obscuring molecular level functional dynamics. We introduce FICTURE2, a statistical model that infers the transcriptional program from which each individual molecule is generated, using gene identity and spatial context alone without segmentation or binning. FICTURE2 learns programs with distinct spatial distributions, ranging from millimeter-scale continuous transition to micron-scale subcellular localization, the coexisting patterns of which then reveal both cellular identity and physiological activity with unified interpretations. FICTURE2 supports both unsupervised and reference-guided inference and is orders of magnitude more efficient than alternative approaches, enabling systematic analysis of large-scale spatial datasets as we demonstrate in data across diverse tissue types profiled by various spatial technologies.

Explore related subjects

genomicsPreprint

Keep this discovery

Explore connections, maps & timelines

BibTeXRIS

Si, Y., Lee, C., Cheng, W., Sweeney, M. D., Yun, J. H., Kim, M., Jun, G., Lee, J. H., Kang, H. M.. 2026-10-30. Molecule-resolved inference reveals the spatial organization of transcriptional programs. https://doi.org/10.64898/2026.10.29.762960

Cite the original work for its findings. Save a collection to share your selection of sources.

KEEP EXPLORING

Related preprints

bioRxiv2026

Somatic variant detection using an individual-specific pangenome reference: a case study of a pediatric acute myeloid leukemia patient

Background: Reference bias can cause false or missed somatic variant calls, also in clinically relevant genes. Pangenomes aim to reduce this bias, but evidence of their diagnostic benefit is lacking. We tested how an individual-specific pangenome affects somatic variant calling in bulk samples and single cells. Results: We built an individual-specific pangenome from Nanopore assemblies of normal hematopoietic stem and progenitor cells of a child with acute myeloid leukemia. We aligned short-read whole-genome sequencing data of bulk and single-cell samples to hg38 and to the pangenome, and called somatic variants with standard downstream tools after anchoring pangenome alignments to hg38. Single nucleotide variants were mostly shared, while indel calls unique to the pangenome were fewer than those unique to hg38, mostly in repeats. Both approaches detected NUP98-NSD1 and FLT3 internal tandem duplication (FLT3-ITD) drivers. The pangenome-based phylogeny showed more parallel branching in comparison to hg38-based tree. The pangenome removed three false-positive PRSS1 variants and, in single cells, frameshift insertion calls next to polyG and polyC repeats. However, surjection to hg38 coordinates split a WT1 driver indel call in a low-complexity region into two separate calls. Validation showed that all somatic structural variant calls, except NUP98-NSD1 and FLT3-ITD, were germline variants or artifacts. Conclusions: The use of individual-specific pangenome led to fewer false-positive indel and driver calls, mainly in repeats and single cells, and changed the branching of the inferred phylogeny. However, surjection artifacts and false-positive somatic structural variants show that pangenome-based results need read-level inspection and validation before clinical use.

genomics↗
bioRxiv2026

Shared genomic signatures of Arctic adaptation among distantly related plant lineages

Extreme environments impose strong selection pressures that can drive convergent adaptation among distantly related species. Arctic plants survive low temperatures, 24-hour light in summer and darkness in winter, short growing seasons, limited precipitation, nutrient-poor soils, and extreme seasonal variation. To test if these conditions are associated with shared genomic signatures, we generated a chromosome-level genome assembly for Oxyria digyna and compared it with existing reference genomes for three distantly related Arctic species (Cochlearia groenlandica, Draba nivalis, Dryas octopetala) and for 16 non-Arctic species from the same families. We identified 214 orthogroups with signatures of positive selection in all Arctic species, 1.6 times the number shared among the non-Arctic species. Although no gene ontology (GO) categories remained significantly enriched after multiple-testing correction, descriptive annotations included genes involved in general functions including protein folding, glycolysis, transcription, and oxidative-stress responses. Orthogroups showing copy number variation were associated with biotic interactions, sexual reproduction, and nutrient acquisition, suggesting that species may respond to biotic and nutrient stresses by shifts in gene copy number. Arctic and alpine genomes contained lower proportions of some repetitive elements, including terminal inverted repeats, and smaller regions of centromeric repeats than non-Arctic genomes. Overall, we see shared genes under positive selection, similar functions of genes with high copy number variation, and lower counts of some repetitive elements in the Arctic plant genomes. Our findings reveal a shared genomic signature of Arctic adaptation across deeply diverged plant lineages, suggesting that evolution in extreme environments repeatedly targets similar genomic features.

genomics↗
bioRxiv2026

MutQA: A Cross-Validated Q&A Dataset for Genetic Mutations

We introduce MutQA, a dataset of 330,387 citation-grounded question-answer (QA) pairs linking genetic mutations to their functional consequences as described in the biomedical literature. Each record pairs a protein-coding mutation with a natural-language question and an evidence-backed answer anchored to sentences in one of 27,442 open-access PubMed Central articles, spanning 6,620 genes and 125,593 gene-variant pairs. The dataset is constructed using CrossValQA, a framework that enforces information isolation between independent LLMs: one model generates QA pairs from an article, a second model answers the same question without access to the first model's response, and a third model judges whether each model's rationale supports the other's answer, so that retained QA pairs are reproducible from the source text rather than hallucinated. Of 401,406 audited QA pairs, 82.7% pass both directions of this cross-validation; those with a recoverable reference protein sequence form the released benchmark. In a 999-pair stratified audit, two expert annotators agree exactly on 86.4% of cross-validated QA pairs (95% CI 83.7-89.0; n = 638), a tier in which about 90% of ratings are "correct". Individual LLM judges agree with individual annotators at a comparable 87.0% on this tier but far less on QA pairs that fail cross-validation. Fine-tuning MutaPLM on MutQA nearly doubles its mean LLM-judge score for question-conditioned mutation-effect answering (1.45 to 2.82 on a 0-5 scale) and improves mutation-recovery accuracy from effect descriptions from 19.4% to 32.1%. Providing variant-matched MutQA retrieval context, without fine-tuning, improves strict answer accuracy across four LLM families by 20.0-28.6 percentage points, demonstrating the framework's ability to improve the reliability of LLM answers. We also apply CrossValQA to proteomics, providing a proof of concept for its use beyond mutation-specific QA. We release the dataset, construction framework, and evaluation code.

genomics↗