Investigating a Microgravity Earth Analog and the Effectiveness of a Nutritional Countermeasure with Machine Learning

Abstract

As space travel advances, it becomes increasingly important to develop effective countermeasures against the negative consequences of microgravity exposure experienced by space travelers, such as bone density loss, muscle atrophy, cardiovascular issues, and the reactivation of dormant viruses. This work presents a machine-learning-based investigation of a nutritional countermeasure designed for and evaluated within a bed rest Earth analog. To improve the effectiveness of the data analysis, multiple transformations and normalization techniques are applied to the original dataset. Various classification models are then employed to assess the efficacy of the countermeasure and to identify physiological and molecular changes corresponding to different phases of the bed rest model. The analysis does not reveal any measurable and consistent effect attributable to the countermeasure. To demonstrate that the modest sample size of the bed rest data does not impose insurmountable limitations on the detectability of potential countermeasure effects, synthetic datasets of varying sizes are also generated and analyzed similarly.