Joint Modeling of Transcriptomic and Morphological Phenotypes for Generative Molecular Design

bioRxiv (Cold Spring Harbor Laboratory) · 2026-02-04

Motivation Phenotypic drug discovery generates rich multi-modal biological data from transcriptomic and morphological measurements, yet translating complex cellular responses into molecular design remains a computational bottleneck. Existing generative methods operate on single modalities and condition on post-treatment measurements without leveraging paired control-treatment dynamics to capture perturbation effects. Results We present Pert2Mol, the first framework for multi-modal phenotype-to-structure generation that integrates transcriptomic and morphological features from paired control-treatment experiments. Pert2Mol employs bidirectional cross-attention between control and treatment states to capture perturbation dynamics, conditioning a rectified flow transformer that generates molecular structures along straight-line trajectories. We introduce Student-Teacher Self-Representation (SERE) learning to stabilize training in high-dimensional multi-modal spaces. On the GDP dataset, Pert2Mol achieves Fréchet ChemNet Distance of 4.996 compared to 7.343 for diffusion baselines and 59.114 for transcriptomics-only methods, while maintaining perfect molecular validity and appropriate physicochemical property distributions. The model demonstrates 84.7% scaffold diversity and 12.4 times faster generation than diffusion approaches with deterministic sampling suitable for hypothesis-driven validation. Availability Code and pretrained models will be available at https://github.com/wangmengbo/Pert2Mol .

Generating Joint Transcriptomic and Morphological Responses to Drug Perturbations via Rectified Flow

bioRxiv (Cold Spring Harbor Laboratory) · 2026-02-03 · 1 citations

Motivation Predicting cellular responses to drug perturbations requires capturing complex dependencies between transcriptomic and morphological changes. Existing approaches model these modalities in isolation, missing critical molecular-phenotypic relationships that occur simultaneously during drug treatment. No current method jointly predicts gene expression profiles and cellular morphology from chemical perturbations. Results We introduce PertFlow, a unified computational frame-work that simultaneously predicts treatment gene expression (bulk RNA-seq) and generates cellular morphology (Cell Painting images) from control cellular states, conditioned on drug metadata. Evaluated on paired RNA-seq and imaging data from 3 cell lines and 40 compounds (17,242 samples), PertFlow achieves Pearson correlation of 0.780±0.264 for transcriptomic prediction and FID of 24.06 for morphological generation. Board-certified pathologists rated generated images with median similarity scores of 7.11-7.89/10. The model successfully recovers known drug mechanisms including microtubule disruption, DNA damage, and MAPK pathway inhibition, with gene enrichment analysis confirming activation of expected biological pathways (EMT, p53, apoptosis). Availability Code and pretrained models will be available at https://github.com/wangmengbo/PertFlow .

Canonical microRNA loss drives tumor development, implicating therapeutic efficacy of enoxacin in angiosarcoma

RNA · 2026-02-26

Angiosarcoma (AS) is a rare and aggressive tumor arising within the endothelium, characterized by a high metastatic rate and poor prognosis. Our prior work established that endothelial loss of Dicer1 , a key enzyme in microRNA (miRNA) processing, drives AS formation in mice, indicating a tumor suppressive role for miRNAs in tumorigenesis. Here, we corroborated this hypothesis by generating a novel conditional knockout model targeting Dgcr8 , a core component of the microprocessor complex required for pri-miRNA processing. Conditional deletion of Dgcr8 phenocopies Dicer1 loss, resulting in spontaneous AS formation and global loss of mature miRNAs. We further demonstrate that treatment with enoxacin (ENX), a repurposed antibiotic known to enhance miRNA processing, reduces viability, migration, and clonogenicity of AS cells. ENX increases the abundance of tumor-suppressive miRNAs and downregulates oncogenic pathways, including pathways related to cell cycle progression, angiogenesis, and cell migration. These results establish the essential role of miRNA biogenesis in suppressing AS and reveal a pharmacologically targetable vulnerability via ENX-mediated enhancement of miRNA expression in tumors.

RestoRect: Degraded Image Restoration via Latent Rectified Flow & Feature Distillation

Open MIND · 2025-01-01

Current approaches for restoration of degraded images face a trade-off: high-performance models are slow for practical use, while fast models produce poor results. Knowledge distillation transfers teacher knowledge to students, but existing static feature matching methods cannot capture how modern transformer architectures dynamically generate features. We propose a novel Latent Rectified Flow Feature Distillation method for restoring degraded images called \textbf{'RestoRect'}. We apply rectified flow to reformulate feature distillation as a generative process where students learn to synthesize teacher-quality features through learnable trajectories in latent space. Our framework combines Retinex decomposition with learnable anisotropic diffusion constraints, and trigonometric color space polarization. We introduce a Feature Layer Extraction loss for robust knowledge transfer between different network architectures through cross-normalized transformer feature alignment with percentile-based outlier detection. RestoRect achieves better training stability, and faster convergence and inference while preserving restoration quality, demonstrating superior results across 15 image restoration datasets, covering 4 tasks, on 10 metrics against baselines.

Canonical microRNA loss drives tumor development implicating therapeutic efficacy of enoxacin in angiosarcoma

bioRxiv (Cold Spring Harbor Laboratory) · 2025-07-17

Abstract Angiosarcoma (AS) is a rare and aggressive tumor arising within the endothelium, characterized by a high metastatic rate and poor prognosis. Our prior work established that endothelial loss of Dicer1 , a key enzyme in microRNA (miRNA) processing, drives AS formation in mice, indicating a tumor suppressive role for miRNAs in tumorigenesis. Here, we corroborated this hypothesis by generating a novel conditional knockout model targeting Dgcr8 , a core component of the microprocessor complex required for pri-miRNA processing. Conditional deletion of Dgcr8 phenocopies Dicer1 loss, resulting in spontaneous AS formation and global loss of mature miRNAs. We further demonstrate that treatment with enoxacin (ENX), a repurposed antibiotic known to enhance miRNA processing, reduces viability, migration, and clonogenicity of AS cells. ENX increases the abundance of tumor-suppressive miRNAs and downregulates oncogenic pathways, including pathways related to cell cycle progression, angiogenesis, and cell migration. These results establish the essential role of miRNA biogenesis in suppressing AS and reveal a pharmacologically targetable vulnerability via ENX-mediated enhancement of miRNA expression in tumors.

Exploring tumor dynamics and responses of prostate cancer to IL-27 based treatment combinations through biodynamic imaging and RNA sequencing analyses

Scientific Reports · 2025-11-26

The tumor microenvironment (TME) is a highly dynamic network shaped by immune, vascular, and stromal interactions, further modulated by extracellular factors. The high tumor heterogeneity complicates treatment due to evolving resistance. Biodynamic imaging (BDI), a 3D coherence-gated holography technique, quantifies intracellular motion to assess phenotypic responses to therapy. We applied BDI to prostate cancer (PCa) treated with IL-27-based combinations to evaluate the dynamic responses of immunotherapy. This study is the first to integrate BDI with RNA-seq to correlate physiological changes with gene expression. Using a subcutaneous PCa model, tumors were treated ex vivo with chemotherapy and immunomodulatory agents. BDI enabled capturing motion frequency shifts (0.01-12.5 Hz), while RNA-seq revealed the molecular changes in the tumors. IL-27 showed potential to modulate the immunologically "cold" TME and enhance therapeutic efficacy. This integrative approach offers a novel platform for evaluating combination strategies in PCa and may inform more effective treatment decisions.

Abstract 1192: Selective sorting of tumor suppressive and oncogenic miRNAs into extracellular vesicles: Implications for cancer progression

Cancer Research · 2025-04-21

Abstract Lung cancer remains the leading cause of cancer-related deaths worldwide, largely due to late diagnoses often coinciding with advanced, metastatic stages. Thus, a deeper understanding of the tumor microenvironment (TME) and the molecular mechanisms driving cancer progression and metastasis is critical. Among the many contributors to the TME, extracellular vesicles (EVs) are crucial in facilitating intercellular communication, enabling the transport of biomolecules such as lipids, RNAs, DNA, and proteins. Of the various biomolecules, microRNAs (miRNAs) have emerged as crucial regulators of both tumor cell behavior and the TME. Indeed, our own research has identified that EVs derived from non-small cell lung cancer (NSCLC) cells carry both tumor-suppressive and oncogenic miRNAs, which exert intrinsic effects on tumor cells and extrinsic effects on the TME. However, the mechanisms governing the selective loading of miRNAs into EVs and their role in modulating TME interactions remain unclear. This study aims to investigate EV-miRNA mediated intercellular communication, with a focus on both selective and non-selective enrichment of specific miRNAs within EVs and their implications in NSCLC. Through small RNA sequencing of EV-derived RNAs from NSCLC cells, we identified enrichment of oncogenic miRNAs, including miR-10b, miR-100, and miR-155, which collectively enhanced the invasive potential of bronchial epithelial cells. Functional antagonism of these miRNAs demonstrated their critical role in driving the invasive and migratory phenotypes induced by EVs. Concurrently, our research revealed a marked downregulation of multiple tumor-suppressive miRNAs in NSCLC cell lines, miRNAs that were selectively enriched within EVs. Sequence alignment analysis identified a conserved motif in approximately 60% of these selectively enriched EV-miRNAs. To determine the functional importance of this motif, we introduced point mutations within the motif that resulted in impaired loading of these miRNAs into EVs and their retention within cancer cells, suggesting that the motif is necessary for this active loading process. Ongoing research seeks to identify key players in the EV-miRNA export pathways and explore the differential loading and expression of miRNAs within EVs in cancer, thereby advancing our understanding of these mechanisms and potentially informing future therapeutic strategies for NSCLC. Citation Format: Samira Piltan, Humna Hasan, Ikjot Sohal, Aadya Pandey, Daniel Urdaneta, Nadia Lanman, Sagar Utturkar, Andrea Kasinski. Selective sorting of tumor suppressive and oncogenic miRNAs into extracellular vesicles: Implications for cancer progression [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2025; Part 1 (Regular Abstracts); 2025 Apr 25-30; Chicago, IL. Philadelphia (PA): AACR; Cancer Res 2025;85(8_Suppl_1):Abstract nr 1192.

Single-nucleus transcriptomics reveal disrupted pathways in the prefrontal cortex of <i>Scn2a</i> -deficient mice

bioRxiv (Cold Spring Harbor Laboratory) · 2025-09-17 · 1 citations

Abstract Truncating variants in SCN2A , which encodes the NaV1.2 sodium channel critical for action potential initiation and propagation, are associated with autism spectrum disorder (ASD) and epilepsy. To investigate SCN2A deficiency–related phenotypes, we developed a preclinical mouse model with under 50% NaV1.2 expression, exhibiting neuronal hyperexcitability and social deficits. However, the neuronal populations and molecular alterations underlying these phenotypes at single-cell resolution have not been investigated. In this study, we conducted single-nucleus RNA sequencing (snRNA-seq) of wild-type (WT), homozygous Scn2a -deficient (HOM) mice, and HOM mice with Scn2a restoration (HOM-FlpO) to examine the effects of Scn2a level on gene expression in the medial prefrontal cortex (mPFC), a critical brain region related to ASD development. Differential expression analysis in GABAergic and glutamatergic neurons between genotypes revealed gene expression enriched in neurotransmitter activity regulation and synapse organization. Lastly, snRNA-seq results in HOM-FlpO identified genes that were rescued after Scn2a restoration. These results reveal that reduced Scn2a expression disrupts RNA transcriptomes in multiple neuronal subtypes, providing insight into cell type–specific mechanisms underlying SCN2A -related disorders.

Abstract 2786: Investigation on the role of tumor suppressive microRNAs in angiosarcoma

Cancer Research · 2025-04-21

Abstract Angiosarcoma (AS) is an aggressive tumor resulting in a very poor prognosis for patients. MicroRNAs (miRNAs) regulate gene expression and play important roles in a variety of diseases including cancer. Our previous studies demonstrate miRNA loss leads to AS in mice. To further understand the role of miRNAs that function as critical tumor suppressors in AS, we performed a miRNA-focused CRIPSR-Cas9 screen. The gRNA library was transduced into a human AS cell line expressing doxycycline (dox)-inducible Cas9. After passaging the cells for 28 days, sgRNA amplicon sequencing was performed to determine the change in the frequency of gRNAs. We anticipated that the loss of function of tumor suppressing miRNAs would lead to gRNA enrichment in the cells. Indeed, three miRNAs were identified as hits with significant enrichment of multiple gRNAs, including miR-200b, miR-181b, and miR-410. We hypothesize these miRNAs act as tumor suppressors and reduce proliferation and cell viability in AS. Based on our RNA-seq data, all three miRNAs are expressed in normal endothelial cells. Additionally, miR-410, miR-181b and a miR-200 family member are significantly downregulated in a human AS cell line and miR-410 is frequently deleted in AS patient samples. We are currently conducting functional validation studies by overexpressing these miRNAs in a panel of AS cell lines by lentiviral based expressing the pre-miRNAs. In preliminary results, the overexpression of miR-410 consistently inhibits cell proliferation and colony formation ability in both human and mouse AS cell lines. Additionally, to gain mechanistic insights for these miRNAs, a novel technique called AgoTRIBE was used to identify important mRNA targets regulated by miRNAs. In AgoTRIBE, the miRNA effector Ago2 is fused to ADAR's RNA editing domain, directing ADAR activity to miRNA natural targets. The ADAR editing events on miRNA targets can be detected by RNA seq. We used a dox-inducible lentiviral vector expressing human ADAR (hADAR) or hADAR-Ago in a mouse AS cell line. In preliminary data, we validated the protein expression of ADAR and ADAR-Ago upon dox treatment and the cellular localization of these complexes in cytoplasm. We are working on validating the function of AgoTRIBE by RNA seq and utilize this tool to identify targets regulated by our screen hits. This project represents the first global screen on miRNAs in AS. Functional validation of the hits from the screen in combination with the mechanistic insights from AgoTRIBE can provide fundamental evidence for the role of tumor suppressive miRNAs in AS. Citation Format: Bozhi Liu, Jillian Stachon, Annaleigh Powell-Benton, Sagar Utturkar, Nimod Janson, Anthony Murphy, Nadia Lanman, Jason Hanna. Investigation on the role of tumor suppressive microRNAs in angiosarcoma [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2025; Part 1 (Regular Abstracts); 2025 Apr 25-30; Chicago, IL. Philadelphia (PA): AACR; Cancer Res 2025;85(8_Suppl_1):Abstract nr 2786.

Autism-associated <i>SCN2A</i> deficiency disrupts cortico-striatal circuitry in human brain assembloids

bioRxiv (Cold Spring Harbor Laboratory) · 2025-06-03 · 2 citations

SUMMARY Profound autism spectrum disorder (ASD) is frequently attributable to single-gene mutations, with SCN2A (voltage-gated sodium channel Na V 1.2) protein-truncating variants (PTVs) being one of the most penetrant. Although cortico-striatal circuitry is implicated as a key node in ASD, the impact of SCN2A deficiency on human neural circuits is unknown. Using the human cortico-striatal assembloid model, we show that the autism-causing PTV SCN2A-C959X impairs long-range cortical axonal projections, reduces striatal spine density, and attenuates excitatory cortical-striatal synaptic transmission. Surprisingly, these assembloids carrying the heterozygous SCN2A nonsense mutation exhibited pronounced network hyperexcitability, a human cell-specific phenotype not observed in Scn2a +/- mice, highlighting a human-specific circuit vulnerability. Collectively, our study unveils human circuit-specific dysfunctions of SCN2A deficiency and SCN2A -mediated ASD. Highlights Axonal projections facilitate synapse formation and functional connectivity in human brain assembloids. Na V 1.2 is expressed along neuronal axons, extending to soma and dendrites in human brain assembloids. SCN2A-C959X disrupts axonal projection patterns, impairs excitatory synaptic transmission, reduces spine density, and results in elevated neuronal excitability. Graphical abstract In brief SCN2A haploinsufficiency impairs cortico-striatal circuitry. SCN2A haploinsufficiency disrupts axon initial segment (AIS) integrity, leading to hyperexcitability (red arrow), reduced axon projections, and impaired synaptic transmission (decreased sEPSCs and altered network firing). These deficits result in dysfunction within the cortico-striatal circuitry.