bioAI Daily Sprouts | 2026-06-09

Search date: 2026-06-09 Window: 2026-06-07 to 2026-06-09. Sources prioritized: Nature, Science, Cell-family, Europe PMC, Crossref, arXiv (collected via the paper-radar skill). No CNS hits this window. Clinical/oncology-only papers without a methods or AI focus were excluded. One false positive dropped (a CsSnI3 perovskite solar cell paper that matched on “cell”).

Papers

1. multiHIVE: Hierarchical Multimodal Deep Generative Modeling for Single-cell Multiomics Research Square (preprint), 2026-06-08. DOI/link Summary: A hierarchical generative model that fuses multiple single-cell omics modalities into a shared, structured latent space. Why it matters: Directly the multimodality-fusion problem for the virtual cell; latent-level fusion is a useful contrast to the Evoformer-style cross-cell attention in Colony. Tags: AI4Bio; single-cell; multiomics; multimodal fusion; generative model

2. Multimodal physical evidence uncovers interpretable gene regulatory networks for perturbation prediction bioRxiv (preprint), 2026-06-07. DOI/link Summary: Uses multimodal evidence to build interpretable gene regulatory networks and predict perturbation responses across the vast perturbation space. Why it matters: Perturbation-response framing overlaps with my spatial-infer grouping/counterfactual logic; interpretability is a bonus. Tags: AI4Bio; GRN; perturbation prediction; multimodal; interpretability

3. Integrating gene regulatory priors into Transformer attention with scTransformer arXiv (preprint), 2026-06-08. DOI/link Summary: Injects gene-regulatory priors into the self-attention mechanism for interpretable scRNA-seq modeling. Why it matters: A concrete recipe for biological priors in attention — relevant to architecture choices for single-cell foundation models. Tags: AI4Bio; single-cell; transformer; attention priors; foundation models

4. Single-cell gene regulatory network reconstruction via a dual-channel fusion graph convolutional network bioRxiv (preprint), 2026-06-07. DOI/link Summary: A dual-channel fusion GCN that reconstructs single-cell GRNs and identifies key regulators. Why it matters: Another fusion-architecture data point (graph + dual-channel) for integrating heterogeneous single-cell signals. Tags: AI4Bio; GRN; graph neural network; fusion; single-cell

5. SpaVGMC: A Unified Representation Learning Framework via Structural and Semantic Alignment in Spatial Transcriptomics Journal of Chemical Information and Modeling, 2026-06-08. DOI/link Summary: Learns spatial-domain representations by aligning structural and semantic views to improve spatial-domain identification. Why it matters: Close to my spatial niche-prediction task; the alignment objective is worth comparing to the Colony niche head. Tags: spatial transcriptomics; representation learning; spatial domains; alignment

6. SNR-ST-Mix: Sample-specific Neighborhood Regression Mixup for Augmented Spatial Transcriptomics Imputation arXiv (preprint), 2026-06-07. DOI/link Summary: A neighborhood-regression Mixup augmentation to recover fine spatial structure from noisy, sparse spatial transcriptomics. Why it matters: Neighborhood-aware augmentation is relevant to the kNN-anchored grouping in my spatial dataset. Tags: spatial transcriptomics; imputation; data augmentation; deep learning

7. Spatial Transcriptomics Open a New Era of Pan-Cancer Analysis (review) Cancer Investigation, 2026-06-07. DOI/link Summary: Reviews how spatial transcriptomics integrates tissue architecture with molecular profiles for pan-cancer analysis. Why it matters: Useful landscape reference for spatial-context modeling. Tags: spatial transcriptomics; pan-cancer; review

Also noted (lower priority)

• Protein/molecule-design transformers: Family-Specialized Transformer for L-cystathionine gamma-lyase (CSBJ, 06-08); Deep Docking part 2 ultra-large virtual screening (Chemical Science, 06-08); MolFoundry structure-aware de novo binder design (ACS Omega, 06-07).
• ~18 disease-specific single-cell / multi-omics signature papers (ccRCC succinylation; NSCLC baicalein 10-gene; HNSCC CDCSI senescence index; IBD / endometriosis / ulcerative colitis microbiome; tonsillar germinal centers in IgA nephropathy; skeletal-muscle development scRNA/ST; pediatric ovary chemotherapy multimodal profiling; uterine gland specification; lung-organoid bone-sarcoma metastasis; etc.) — excluded as clinical-only.