Discovering novel phenotypes with automatically inferred dynamic models: a partial melanocyte conversion in Xenopus.

Area of Science:

• * Developmental Biology
• * Computational Biology
• * Regenerative Medicine

Background:

• * Cellular control networks are crucial for regenerative medicine, requiring models to predict outcomes of perturbations.
• * Previous studies observed an all-or-none conversion of melanocytes to a metastatic-like phenotype in Xenopus.
• * Existing models struggled to explain the stochastic, all-or-none nature of this cellular conversion.

Purpose of the Study:

• * To develop a machine learning method to model the all-or-none melanocyte conversion.
• * To predict novel perturbations that could generate a partial conversion phenotype.
• * To validate the model's predictions through in vivo experiments.

Main Methods:

• * Developed a machine learning approach to infer a dynamic model from stochastic, all-or-none data.
• * Utilized in silico perturbations to explore potential manipulations of the cellular network.
• * Conducted in vivo experiments applying predicted drug combinations (altanserin, reserpine, VP16-XlCreb1) in Xenopus.

Main Results:

• * Successfully inferred a model explaining the all-or-none melanocyte conversion.
• * Predicted that a specific combination of three reagents would induce partial conversion.
• * In vivo application of the predicted combination resulted in the novel partial melanocyte conversion phenotype.

Conclusions:

• * Automated analysis of dynamic signaling network models can discover novel phenotypes.
• * Predictive modeling enables precise identification of manipulations to achieve desired cellular outcomes.
• * This approach holds significant potential for advancing regenerative medicine and drug discovery.