OmniCellTOSG: The First Cell Text-Omic Signaling Graphs Dataset for Joint LLM and GNN Modeling

Abstract: Complex cell signaling systems -- governed by varying protein abundances and interactions -- generate diverse cell types across organs. These systems evolve under influences such as age, sex, diet, environmental exposures, and diseases, making them challenging to decode given the involvement of tens of thousands of genes and proteins. Recently, hundreds of millions of single-cell omics data have provided a robust foundation for understanding these signaling networks within various cell subpopulations and conditions. Inspired by the success of large foundation models (for example, LLMs and large vision models) pre-trained on massive datasets, we introduce OmniCellTOSG, the first dataset of cell text-omic signaling graphs (TOSGs). Each TOSG represents the signaling network of an individual or meta-cell and is labeled with information such as organ, disease, sex, age, and cell subtype. OmniCellTOSG offers two key contributions. First, it introduces a novel graph model that integrates human-readable annotations -- such as biological functions, cellular locations, signaling pathways, related diseases, and drugs -- with quantitative gene and protein abundance data, enabling graph reasoning to decode cell signaling. This approach calls for new joint models combining LLMs and graph neural networks. Second, the dataset is built from single-cell RNA sequencing data of approximately 120 million cells from diverse tissues and conditions (healthy and diseased) and is fully compatible with PyTorch. This facilitates the development of innovative cell signaling models that could transform research in life sciences, healthcare, and precision medicine. The OmniCellTOSG dataset is continuously expanding and will be updated regularly. The dataset and code are available at https://github.com/FuhaiLiAiLab/OmniCellTOSG.

OmniCellTOSG: A Dataset for Integrated Cellular Signaling Analysis

The paper "OmniCellTOSG: The First Cell Text-Omic Signaling Graphs Dataset for Joint LLM and GNN Modeling" introduces a novel and comprehensive dataset named OmniCellTOSG. This dataset represents a pioneering effort to facilitate the integration of LLMs with graph neural networks (GNNs) for the study of cell signaling systems. Given the complexity of cell signaling systems which involve intricate protein interactions affecting diverse biological functions, the OmniCellTOSG serves as a powerful resource for decoding these systems under various physiological and pathological conditions.

Contributions of OmniCellTOSG

The OmniCellTOSG dataset introduces the concept of Text-Omic Signaling Graphs (TOSGs), which uniquely integrate human-readable annotations with quantitative omic data, representing signaling interactions within individual cells. This integration is achieved through the following key steps:

1. Data Generation and Integration: The dataset incorporates single-cell RNA sequencing (scRNAseq) data from approximately 120 million cells across various organs and conditions, including health and disease states. The TOSGs represent each cell or cellular subtype with associated metadata such as organ type, disease state, age, and sex.
2. Graph Representation: Each TOSG is a graph data model that combines textual annotations—providing biological functions, cell locations, and associated pathways—with numeric features from gene/protein expression levels. This hybrid model allows for more nuanced graph reasoning and inference using both LLMs and GNNs.
3. Data Compatibility: The dataset is structured for compatibility with PyTorch, enabling easy integration and use in machine learning pipelines. The OmniCellTOSG's ongoing updates will expand its usability and relevance as a tool for continuous research.

Implications and Future Prospects

Theoretical Implications:

The integration of LLMs and GNNs using TOSGs represents a novel direction in AI-driven bioinformatics research. By utilizing vast textual and numeric datasets, the OmniCellTOSG supports the development of foundation models that capture the intricate patterns in cellular signaling. These models could provide a more generalized understanding of cell signaling that is less prone to the overfitting issues of disease-specific approaches.

Practical Implications:

The resource promises to advance the field of precision medicine by facilitating the identification of complex signaling pathways and potential therapeutic targets in diseases like cancer and Alzheimer's disease. Biomedical researchers can explore the effectiveness of drug interventions and predict outcomes by simulating cellular alterations in signaling pathways.

Future Developments:

The paper suggests continued extension of the OmniCellTOSG dataset to incorporate a broader array of conditions, diseases, and demographic factors. This expansion will enhance its capability to represent diverse biological signals accurately. Additionally, the integration of advanced LLMs to extract richer textual knowledge from biomedical literature could substantially enrich the dataset.

Conclusion

The OmniCellTOSG dataset marks an important step toward the integration of comprehensive omic data with advanced AI techniques, offering a robust platform for both theoretical explorations and practical applications in understanding cell signaling mechanisms. It provides a vital base for researchers aiming to develop innovative computational models that can handle the complexity of biological systems in dynamic and diverse environments. As its scope grows, OmniCellTOSG is poised to facilitate significant breakthroughs in life sciences and precision medicine research.