
Prof. Kumar Selvarajoo
Key Laboratory of Preclinical Research of Antitumor Drugs, Taizhou Institute of Zhejiang University, China
Title: From Systems Biology to Systems AI: Decoding Cancer Heterogeneity for Predictive Drug Discovery
Abstract:
Cancer is a complex and dynamic disease characterized by
extensive molecular heterogeneity, cellular plasticity, and adaptive responses
to therapeutic interventions. Despite remarkable advances in high-throughput
omics technologies and artificial intelligence (AI), our ability to translate
molecular information into reliable therapeutic predictions remains limited. A
fundamental challenge is that conventional data-driven approaches predominantly
identify statistical associations and differentially expressed genes, often
without revealing the underlying regulatory mechanisms that govern cellular
behavior. Addressing this challenge requires a shift from descriptive molecular
profiling toward mechanistic and predictive understanding of biological
systems.
In this keynote, I will discuss how our research has
evolved from mechanistic systems biology to the emerging concept of Systems AI,
integrating mathematical modeling, stochastic transcriptional regulation, and
multimodal machine learning to understand cancer heterogeneity and identify
therapeutic opportunities.
I will first highlight how mechanistic modeling can uncover
therapeutic vulnerabilities that may be overlooked by conventional experimental
approaches. Using tumor necrosis factor-related apoptosis-inducing ligand
(TRAIL) resistance as an example, our systems-level computational analyses
identified protein kinase C (PKC) signaling as a potential therapeutic
intervention point. Subsequent experimental investigations established the
involvement of PKCδ in regulating TRAIL sensitivity, illustrating the value of
integrating computational predictions with experimental validation to uncover
actionable biological mechanisms.
I will then explore how stochastic gene expression
contributes to cellular heterogeneity and potentially influences cancer
progression. Our investigations of chronic lymphocytic leukemia revealed
widespread switch-like gene-expression patterns, termed toggle genes, characterized
by distinct ON and OFF expression states. These observations suggest that
transcriptional variability and cellular state transitions contain biologically
meaningful information that is frequently obscured by population-averaged
measurements. Building on this concept, our recent work developed a generalized
stochastic transcriptional regulation framework capable of reconstructing
ovarian cancer transcriptomes through process-driven simulations. By examining
transcriptional and RNA degradation parameters, this approach provides an
alternative perspective on the regulatory processes underlying gene-expression
differences and variability between normal and malignant tissues.
Beyond mechanistic modeling, I will discuss how integrating
diverse molecular and clinical information through AI can improve the
interpretation and prediction of therapeutic responses. Our multimodal
machine-learning framework, MOMLIN, combines genomic, transcriptomic, clinical,
and tumor microenvironment information to identify interpretable biomarkers
associated with drug responses in breast cancer. This illustrates the
complementary strengths of mechanistic and data-driven approaches in advancing
precision oncology.
Finally, I will introduce the broader vision of Systems AI,
in which mechanistic models, stochastic simulations, multimodal learning, and experimental
perturbations are integrated into an iterative framework for biological
discovery. Rather than relying exclusively on statistical correlations, Systems
AI seeks to generate biologically interpretable, experimentally testable
predictions of cellular behavior and therapeutic responses. Such integration
may ultimately support the development of predictive biological digital twins
for investigating disease mechanisms, evaluating treatment strategies, and
prioritizing therapeutic targets.
Together, these advances highlight a transition from
observing molecular changes to understanding, predicting, and potentially
controlling complex biological systems, opening new opportunities for
mechanistically informed precision medicine and next-generation drug discovery.
Biography:
Kumar
Selvarajoo is Chief Scientist and Professor at the Taizhou Institute of
Zhejiang University, China, where he leads research integrating systems
biology, computational biology, and artificial intelligence (AI) for
understanding complex biological systems and advancing drug discovery.
Previously, he served as a Senior Principal Investigator at the Bioinformatics
Institute, Agency for Science, Technology and Research (A*STAR), Singapore, and
Associate Professor of Systems Immunology at Keio University, Japan. He has
also held adjunct faculty appointments at the National University of Singapore
and Nanyang Technological University.
His
research focuses on mechanistic modeling, stochastic gene regulation,
multi-omics integration, and AI-driven approaches to cancer biology,
immunology, and precision medicine. His work bridges mathematical and
computational predictions with experimental investigations to uncover
biological mechanisms and identify potential therapeutic targets. More
recently, he has been developing the concept of Systems AI, which integrates
mechanistic systems biology with machine learning to enable interpretable and
predictive biological discoveries.
Professor
Selvarajoo has authored over 100 scientific publications and the Springer
book Immuno Systems Biology. He serves on the editorial boards of
several international scientific journals and has contributed extensively to
interdisciplinary research at the interface of biology, mathematics, and AI.