Objectives: This study aims to demonstrate the utility of Artificial Intelligence in enhancing the reproducibility of mathematical disease models. Through a case study of a disease progression mechanistic model of Retinitis Pigmentosa [1], the potential of AI in streamlining the model re-implementation process is showcased.
Methods: Chat GPT 4, a sophisticated large language model developed by OpenAI, customized by Mathpix with additional functionalities, was employed. This customized model enabled the extraction of equations from PDF documents and their conversion into Antimony, a human-readable, text-based language. Subsequently, the model was translated into Systems Biology Markup Language (SBML) using the Python package tellurium. The converted model was then uploaded onto the jinkō platform for simulations.
Results: The efficient and accurate implementation of the RP model using AI techniques was demonstrated in this study. The disease progression biological phenomena such as the evolutions of rods, cones and the nutrient pool (representative of the total number of retinal-pigment epithelium neuroprotective factors, growth factors and nutrients) as reported in the literature was successfully reproduced.
Conclusion: The findings of this study underscore the value of AI in enhancing the reproducibility of mathematical disease models. By automating the re-implementation process and facilitating swift verification, AI offers a promising avenue for decision-makers to independently validate model results. This encourages the broader adoption of AI in mathematical model re-implementations to ensure reliability and reproducibility of the results.
Citations: [1] Camacho, E. T., Punzo, C. & Wirkus, S. A. Quantifying the metabolic contribution to photoreceptor death in retinitis pigmentosa via a mathematical model. Journal of Theoretical Biology vol. 408 75–87 (2016).
