Merrell, A. | Search bioRxiv

bioRxiv2025

DINOcell: Learning Generalizable Perturbation Effects through Self-Distillation

AO_SCPLOWBSTRACTC_SCPLOWPredicting cellular responses to therapeutics is a promising approach for novel target discovery. However, state-of-the-art computational models designed to predict perturbation effects struggle to generalize and outperform simple baselines. We present DINOcell, a weakly supervised framework that adapts self-distillation to single-cell transcriptomics for predicting perturbation effects. We demonstrate that DINOcell outperforms baselines in predicting the effects of single gene perturbations. Furthermore, DINOcell accurately predicts non-additive effects of combination perturbations, indicating its capacity to model complex genetic interactions. Finally, we show that DINOcell learns representations that capture biological signals and is a promising, generalizable approach for in silico perturbation modeling, providing a valuable tool for accelerating therapeutic target discovery.

bioinformatics↗

bioRxiv2020

Dissecting phenotypic transitions in metastatic disease via photoconversion-based isolation

Cancer patients presenting with surgically resectable disease often harbor occult metastases, a potential source of relapse that is targetable only through systemic therapy. Studies of this occult fraction have been limited by a lack of tools with which to isolate discrete cells based on spatial grounds. We developed PIC-IT, photoconversion-based isolation technique allowing efficient recovery of cell clusters of any size including solitary disseminated tumor cells (DTCs), which are largely inaccessible otherwise. In a murine pancreatic cancer model, transcriptional profiling of spontaneously arising DTCs revealed phenotypic heterogeneity, functionally reduced propensity to proliferate and enrichment for inflammatory-response phenotype associated with NF-{kappa}B /AP-1 signaling. Pharmacological inhibition of NF-{kappa}B depleted DTCs but had no effect on macrometastases, suggesting DTCs are particularly dependent on this pathway. PIC-IT enables systematic investigation of the earliest stages of metastatic colonization. Moreover, this new technique can be applied to other biological systems in which isolation and characterization of spatially distinct cell populations is not currently feasible.

cancer biology↗