scDisentangle: sequential disentanglement for single-cell perturbation response prediction
Predicting the transcriptional response of single cells to perturbation is a central goal of single-cell modeling, with broad application to analysis and experimental design. Existing methods often fail to beat simple baselines and offer little insight into how biological variation is organized. We present scDisentangle, a model that decomposes single-cell expression into observed attributes and…
A new model named scDisentangle has been developed to predict the transcriptional response of single cells to perturbation. This model aims to enhance single-cell modeling, which has broad applications in analysis and experimental design. However, existing methods often fail to surpass simple baselines and provide limited insight into the organization of biological variation.
scDisentangle addresses these challenges by decomposing single-cell expression into observed attributes and a sequence of residual components. This decomposition generates counterfactuals that preserve the identity of the source cell. The model has shown superior performance compared to current models and naive baselines in various perturbation scenarios. These scenarios include cytokine stimulation, liver infection, and unseen combinatorial genetic perturbations.
Accuracy with scDisentangle was largely independent of how closely a response resembled the average perturbation effect, indicating its robustness. In a specific study involving an immune checkpoint inhibitor-associated myocarditis cohort, scDisentangle predicted donor-specific intracardiac CD8+ T cell states from peripheral blood. This ability to predict cell states from peripheral blood could have significant implications for understanding and treating various diseases.
Furthermore, in a castration-regeneration prostate atlas, scDisentangle's predictions recapitulated luminal dynamics and progenitor-marker kinetics upon androgen add-back. This suggests that scDisentangle could also be useful in understanding and predicting biological processes in other complex systems.
The model's coarse-to-fine latent structure also aligns with hematopoietic lineage organization. Early levels of the latent structure capture a less differentiated core, which is progressively refined into cell-type-specific programs as the model progresses. This fine-grained resolution of cell identity and differentiation could provide valuable insights into the complex processes governed by single-cell biology.
Written by urgent.news from bioRxiv's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
