The inductive (data-based) and the deductive (hypothesis-based) approaches have played a complementary and mutually beneficial role in the history of science, whereby observations have led to the postulation of hypotheses that are subsequently tested by properly designed experiments. This forms an evolutionary process of hypothesis formulation and testing, leading to scientific advancement. In life sciences and medicine, the importance of discovering and quantifying the physiological mechanisms under normal and pathological conditions has given rise to mechanism-based modeling methods (e.g. compartmental or structural modeling) which rely on the current state of understanding of the system under study. However, the intrinsic complexity of physiological systems and the need for validation of the structural models present formidable challenges in the mechanism-based approach and motivate the complementary use of data-based modeling approaches (typically input-output or stimulus-response formulations). The latter seek to capture the essential functional characteristics of the physiological system in a manner consistent with the available data. Subsequent analysis of the obtained data-based models suggest hypothesis-based model forms that encapsulate the relevant physiological mechanisms and can be tested through properly designed experiments. In this process, the data-based model serves as “ground truth” for the validity of an equivalent hypothesis-based or mechanism-based model. Our experience over the last 30 years shows that this “virtuous cycle” of model development is enabled by the synergistic use of data-based and hypothesis-based approaches.

The study of functional and structural complexity in living systems requires reliable and robust modeling tools in a hierarchical context of multiple scales of time and space. Although mechanism-based models remain the ultimate objective of multi-scale modeling, data-based models can be helpful in pursuing this goal because of their applicability to arbitrary levels of systemic organization from molecular to cellular to multi-cellular to organ to multi-organ etc. This broad applicability depends on appropriate methods of modeling/analysis within the constraints imposed by experimental limitations. This talk seeks to stimulate our thinking on the synergistic use of data-based and hypothesis-based modeling methods in a practical context. It will summarize our findings to date and will present illustrative examples from neural and metabolic systems where this synergistic approach has yielded useful insights.