Spacia: Unveiling a New Era of Cell-Cell Communication in Spatial Transcriptomics
Dr. Tao Wang
Dr. Tao Wang runs a top computational immunology lab in UT Southwestern Medical Center. Machine learning, statistics, medicine, and biology are the four integral pillars of his interdisciplinary research program. Dr. Wang has been working on mining public and in-house high throughput data to achieve a deeper understanding of the mechanisms of various human diseases, with a heavy emphasis on immunological diseases, and their diagnosis, prognosis, and treatment. These approaches include, for example, modeling of T/B cell antigens and T/B cell receptor sequences, profiling immune cell infiltrates in bulk tissues and analyses of single cell profiling (scRNA-seq, spatially resolved transcriptomics, and CyTOF) data, and integrative analyses of genomics data with electronic medical record data. The ultimate goal of Wang lab is to impact the prognosis and treatment of patients suffering from various diseases, through modeling of high dimensional data, especially genomics data.
Dr. Yang Xie
Dr. Yang Xie holds the Raymond D. and Patsy R. Nasher Distinguished Chair in Cancer Research and is the Associate Dean for Data Sciences at UT Southwestern Medical Center. She is the founding director of the Quantitative Biomedical Research Center (QBRC), the Pediatric Cancer Data Commons (PCDC), and the Cancer Center Data Science Shared Resources (DSSR) at the Harold C. Simmons Comprehensive Cancer Center. Dr. Yang Xie received her training in biostatistics, medicine and epidemiology. Her research lab focuses on medical informatics, developing predictive and prognostic biomarkers, and precision medicine. She is currently the PI of an NIH Maximizing Investigators' Research Award (MIRA) grant, MPI of an NIAID U01 grant and PI of the Pediatric Cancer Data Core at CPRIT.
Dr. Xinlei (Sherry) Wang
Dr. Xinlei (Sherry) Wang is Jenkins-Garratt Endowed Professor of Statistics and Data Science and Director for Research, Division of Data Science at UTA. Her research interests include Bayesian statistics, statistical omics, machine/deep learning and AI, integrative/meta-analysis, and order-related sampling design, theory, and inference. Sher has rich experience in developing and applying statistical and computational methods for analysis of complex and diverse data. She leverages this expertise to extensively collaborate with researchers across various fields, including biomedicine, health care, business, computer science, engineering, chemistry, physics, and mathematics. Her research projects have been generously supported by external funding, including NCI R01, NIGMS R15, NIDA R21 and R33, NSF-DMS grants as PI or subcontract PI. In 2020, she was honored with the Gerald J. Ford Senior Research Fellowship at SMU. In 2024, She was elected as an ASA fellow.
Regarding the research background and significance, does this work discover new knowledge or solve existing problems within the field? Please elaborate in detail.
Spacia uses a novel approach to infer cell-to-cell communication (CCC) using data from spatially resolved transcriptomics technologies (SRTs). This approach improves upon the capabilities of current methods by achieving single-cell resolution, not requiring prior knowledge of interaction pathways, and accounting for complex, multiple-to-one and one-to-multiple relationships.
How did the reviewers evaluate (praise) it?
The reviewers were impressed by Spacia’s novel approach and additional capabilities over existing methods.
If this achievement has potential applications, what are some specific applications it might have in a few years?
We hope that Spacia can help researchers dissect the complex interactions occurring within heterogeneous cellular environments, such as tumors. It will enable the discovery and understanding of novel CCCs and their roles in normal and disease biological processes. As the cost of SRTs decline over time, Spacia may be used in a more translational setting to identify potential drug targets.
Can you recount the specific steps or stages from setting the research topic to the successful completion of the research?
With the appearance of single-cell resolution SRTs, we wanted to take advantage of the spatial information offered by these techniques to improve CCC inference. At the same time, we also wanted to take a different approach. Instead of limiting CCCs to ligand-receptor interactions, we used what we refer to as a multiple-instance learning framework to simultaneously model and solve the complex interactions among cells. We first fine-tuned the model on simulated data as a proof of concept, then used Spacia to analyze real SRT datasets from different technologies. Using a combination of existing knowledge and other types of transcriptomics data we generated or collected from public sources, we confirmed Spacia’s findings and found them to be overall generalizable. We also discovered some clinically relevant and biologically interesting findings through our analysis. Throughout this process, we had to overcome many different issues in optimizing and refining the model, finding and processing diverse types of data, and analyzing and interpreting the results.
Were there any memorable events during the research? You can tell a story about anything related to people, events, or objects.
While significant amounts of resources and time have been spent by the field on developing and applying spatially resolved transcriptomics, few works have established the true value of these technologies to reveal insights that cannot be efficiently provided by traditional experimental assays, bulk sequencing or single cell sequencing. I have several collaborators who expressed to me that they did not feel the current hype of spatial transcriptomics and even single cell sequencing were helping them in any way, because their scientific questions could all be addressed by bulk sequencing or even traditional IHC already. While there are certain aspects of their statement that are valid, we here showed that SRT data offer much more than just showing the spatial distribution of cells and their cell types in a tissue sample. We can actually perform more complicated analyses that reveal very novel interesting new biology from these data. We plan to further dive into SRT data to demonstrate all the exciting discoveries they can make.
Is there a follow-up plan based on this research? If so, please elaborate.
We are looking to follow up on some of the novel and biologically interesting findings we made during our analysis. There are also aspects of Spacia, namely speed and ease of use that we would like to continue to improve. In this respect, we hope to gather input from other interested users to gain a better understanding of their needs. We also hope that other researchers can adapt and expand our multiple-instance learning framework to other problems and fields.
Without a doubt, AI is one of the hot topics of 2024, requiring extensive data support in its development. What assistance can biostatistics offer to the development of AI?
We believe that biostatistics is a critical part of AI, especially in data processing and analysis of results to arrive at more profound understandings. In addition, statistical learning methods, such as what we have used in this paper, have the potential to offer elegant solutions with advantages in interpretability and stability. We believe there is a bright future for biostatistics in the age of AI.
Proofread by: Hongtu Zhu