Multi-Modal and Multi-Attribute Generation of Single Cells with CFGen

2024-07-16

AI Summary

Existing single-cell generative models predominantly rely on continuous expression approximations, neglecting the intrinsic discrete-count nature of RNA-seq data, and lack support for controllable generation across multimodal (e.g., RNA+ATAC) and multi-attribute (e.g., cell type, batch, state) conditions. To address this, we propose CFGen—the first conditional flow-matching generative model specifically designed for discrete single-cell count data. CFGen directly models raw UMI counts via conditional normalizing flows, integrating multimodal alignment embeddings and batch-aware conditional encodings to enable fine-grained, interpretable cross-modal joint generation. Evaluated on multiple real-world datasets, CFGen significantly improves rare cell type augmentation, batch-effect correction, and developmental trajectory simulation. Moreover, it faithfully recapitulates biologically critical features—including gene co-expression patterns and bimodal expression distributions—demonstrating superior fidelity and biological plausibility.

Technology Category

• Natural Language Processing: Generation
• Machine Learning: Deep Generative Models & Autoencoders
• Computer Vision: Generative Adversarial Networks (GANs) for Vision

Application Category

• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG
• Social Networks and Social Media: Generative AI / large language models and their impact on social systems
• Economics, Online Markets and Human Computation: Economic ramifications for generative AI infrastructure and applications

Abstract

Generative modeling of single-cell RNA-seq data is crucial for tasks like trajectory inference, batch effect removal, and simulation of realistic cellular data. However, recent deep generative models simulating synthetic single cells from noise operate on pre-processed continuous gene expression approximations, overlooking the discrete nature of single-cell data, which limits their effectiveness and hinders the incorporation of robust noise models. Additionally, aspects like controllable multi-modal and multi-label generation of cellular data remain underexplored. This work introduces CellFlow for Generation (CFGen), a flow-based conditional generative model that preserves the inherent discreteness of single-cell data. CFGen generates whole-genome multi-modal single-cell data reliably, improving the recovery of crucial biological data characteristics while tackling relevant generative tasks such as rare cell type augmentation and batch correction. We also introduce a novel framework for compositional data generation using Flow Matching. By showcasing CFGen on a diverse set of biological datasets and settings, we provide evidence of its value to the fields of computational biology and deep generative models.

Problem

• Generative modeling of single-cell RNA-seq data for biological tasks.
• Overcoming limitations of deep generative models with discrete data.
• Enabling controllable multi-modal and multi-label cellular data generation.

Innovation

• CFGen preserves single-cell data discreteness
• Generates multi-modal, multi-label cellular data
• Uses Flow Matching for compositional data generation