Computer Assisted Deductive Integration

A complementary approach between heuristic and mathematical modeling since, in biology, functional linearity cannot be assumed and the components of a model cannot incorporate solely what is known.

Heuristic modeling plays the role of an architect (defines the nature, the structure, the functionalities and the contextual constraints of the system under study) whereas mathematical modeling, to be implemented at a later stage, plays the role of an engineer (reveals the dynamics and robustness of the structures within the system while defining the set of parameters sufficient to give rise to similar or very different phenotypes).


CADI discovery is a problems solving approach evaluating each step in a process, searching for satisfactory solutions rather than for optimal solutions, using all available qualitative information instead of quantitative information. 

Since it strictly relies upon strict and systematic implementation of negative selection of hypotheses, models arising from this procedure contain elements that had hitherto never been described but cannot be refuted by current knowledge and/or available biological data, thereby generating novel understanding.

Screenshot 2019-12-01 at 17.53.16.png

Diagram depicting the model-building procedure. The circular shape of the overall process, embodied by labelled black arrows, indicates the iterative nature of the entire analytical procedure where each main analytical step is represented by a labelled disc. Since a circle does not have a defined beginning, the model-building procedure may be initiated from any given step in the process. The dotted red arrows indicate the internal iterative loops within the overall process. The go board on the lower left (black & white tokens on a checker board) represents the distribution of untested working hypotheses directly generated from the literature while its lower right counterpart represents the distribution of hypotheses that have resisted all destruction attempts. These undestroyed hypotheses are then merged to produce interaction maps that are in turn merged to produce a hypothetical physiological model. During each of these two steps, numerous novel working hypotheses suggesting hitherto undescribed biological events are being generated. These are, in turn, subjected to the iterative negative (destructive) selection procedure. This is represented by dotted red arrows linking ‘interaction maps’ and ‘hypothetical physiological model’ (lower and upper right discs) with ‘identified events’ and ‘production of working hypotheses’ (upper and lower left discs). Hence, this model-building process involves multiple levels of internal crosscheck procedures designed to eliminate any hypothesis that is not directly as well as indirectly supported by multiple data intersects (scientific literature and publicly accessible databases). The model resulting from this process is then subjected to direct experimental evaluation and the new data thus generated can be in turn integrated into the model, correcting errors and misdirections.