Tsang Lab @ Yale - Research

Projects involve systems, computational, and quantitative immunology, which motivate both approaches and questions for understanding the immune system as a system interacting with physiology. Our work often involve multimodal immune profiling, machine learning, dynamical modeling, and experiments using ex vivo/in vitro/animal models. Some ongoing projects are highlighted below. We are open to new ideas in the broad realm of systems and engineering immunology, including new biological questions and computational/technological approaches.

Human immune variation, immune setpoint, and immune health

We have a long-standing interest in understanding the molecular and cellular basis of immune variability in the human population. In addition to utilizing exposures such as infections and natural variations including genetic and exposure histories, we utilize vaccines as ethical, timed perturbations to assess the immune system of diverse populations (Sparks, Rachmaninoff, Lau, Hirsch et al Nature Medicine 2024; Mulè et al Immunity 2024; Sparks, Lau, Liu et al Nature 2023; Tsang Trends in Immunology 2015; Cheung et al eLife 2023; Liu, Martins, Lau, Rachmaninoff Cell 2021; Kotliarov, Sparks et al Nature Medicine 2020; Tsang et al Trends in Immunology 2020; HIPC-CHI Consortium et al Science Immunology 2017; Tsang et al Cell 2014).

We often design and conduct our own human subject studies, from human study and sample collection protocol development to data generation, analysis, and predictive and quantitative model development. For example, see Sparks, Rachmaninoff, Lau, Hirsch et al Nature Medicine 2024; Mulè et al Immunity 2024; Sparks, Lau, Liu et al Nature 2023; Tsang et al Cell 2014.

We also collaborate extensively with colleagues and consortia with existing human samples and data. For example, by generating and combining data from diverse human cohorts, including cross-sectional, longitudinal, and household studies, we recently launched the Flu Diversity Project together with Sarah Cobey, Ben Cowling and colleagues. We seek to understand how vaccines and prior exposures shape personal immune status in both antigen-specific and -agnostic ways, and why influenza vaccine and infection responses are so heterogeneous across individuals and populations.

A related issue we are working on is vaccine hypo-responsiveness, which has been recognized as a major roadblock for vaccine efficacy for some populations. For example, experimental malaria vaccines like PfSPZ are generally known to have high efficacy in US and EU based trials, but their protection efficacy drops significantly (including in children) in endemic regions of Africa. We have ongoing projects in collaboration with Maria Yazdanbakhsh, Claudia Daubenberger, Steve Hoffman, Carlota Dobaño, Gemma Moncunill and colleagues, in which we use systems immunology and quantitative modeling approaches to study how baseline immune status and set points differ across geographic regions, and how those differences may explain malaria vaccine hypo-responsiveness. This research could illuminate novel strategies to modulate baseline immune system states to "restore" vaccine efficacy.

Tissue inflammation and homeostasis

We are interested in understanding, in quantitative and network biology terms, how immune cells traffick to tissues and how homeostasis is maintained and deviations from homeostasis is detected. We are a part of the CZI Single Cell Inflammation Program and by using skin as a model, we aim to answer some of these questions in humans. We are also seeking to develop complementary animal and organoid models to quantify and model tissue dynamics.

Systems immunology of maternal-infant dyads

To investigate the origin and development of individuality and personal immune states, we trace immune development and vaccine response starting from pregnancy using multimodal immune profiling, single cell analysis, and systems serology; we monitor and quantitative model vaccine responses in pregnant moms and their infants. Similarly, together with Eva Harris and colleagues, we study immune development in Nicaraguan children by longitudinally following them from infancy to adolescence, e.g., to decipher how immune status and set points are established in humans.

Immune memory

We seek to understand both antigen-agnostic and antigen-specific memory, determinants of their memorization timescales, and why some vaccines (e.g., yellow fever) can induce ultra-long lasting protection and immune memory (even with only one dose of the vaccine) while others, like COVID-19 mRNA vaccines, seem to provide less durable protection. What are the early response predictors of durability and memory? How can we program the immune system for long-lasting durable memory responses on both the innate and adaptive sides? We are integrating animal models, human studies, and multiomics single cell longitudinal analyses and computational modeling to address these issues.

Methodological research

We are broadly interested in developing, refining, and scaling up computational and experimental methods to enable systems immunology. We are also broadly interested in studying the "design principles" of the immune system and immune responses.

For example, to enable multimodal profiling and monitoring of human immune states in populations over time (e.g., before and after vaccination) and to develop predictive models and identify predictors and determinants of immune response outcomes, we have developed and scaled up approaches for sample-multiplexed, multimodal single cell analysis (e.g., see Mulè et al Immunity 2024, Liu, Martins, Lau, and Rachmaninoff et al Cell 2021, and Sparks, Lau, Liu et al Nature 2023). These include computational denoising and normalization methods (e.g., see Mulè, Martins, Tsang Nat. Comm. 2022), barcoding schemes combining genetics and hashtag multiplexing (e.g., enables pooled and sort approaches to enrich for rare cell populations), a machine learning toolkit/R package to identify predictors of responses (Candia and Tsang, BMC Bioinformatics 2020), shallow sequencing to develop machine learning predictors of cancer outcomes (Milanez-Almeida et al Nature Medicine 2020), and an approach to analyze single cell stimulation responses (Farmer et al Biorxiv 2022) . We have also been integrating dynamical/stochastic mechanistic modeling and machine learning to achieve fast prediction of emergent phenotypes and intuitive/interpretable understanding of the key determinants of the phenotypes (see Park et al. Biorxiv 2019, Martins and Narayanan et al Cell Systems 2017 and Wong et al. Cell 2021).

Our lab developed a web-based, crowdsourcing platform (OMiCC) for the management, reuse and meta-analysis of large-scale public data sets; OMiCC enables scientists without specialized training to utilize large-scale data from multiple studies to generate and test biological hypotheses (Sparks et al Nature Biotech 2016 and Liu et al STAR Protocols 2022). We illustrated, through a crowdsourcing experiment involving NIH volunteer scientists, how OMiCC can enable a group of non-computational biologists to utilize publicly available gene expression data to construct a multi-study “virtual” dataset of autoimmune diseases in both humans and animal models followed by meta-analysis to uncover disease signatures (Sparks et al Immunity 2016 and Lau et al F1000 Research 2016).