Optimizing antiviral therapy for COVID-19 with learned pathogenic model

COVID-19 together with variants have caused an unprecedented amount of mental and economic turmoil with ever increasing fatality and no proven therapies in sight. The healthcare industry is racing to find a cure with multitude of clinical trials underway to access the efficacy of repurposed antivirals, however the much needed insights into the dynamics of pathogenesis of SARS-CoV-2 and corresponding pharmacology of antivirals are lacking. This paper introduces systematic pathological model learning of COVID-19 dynamics followed by derivative free optimization based multi objective drug rescheduling. The pathological model learnt from clinical data of severe COVID-19 patients treated with remdesivir could additionally predict immune T cells response and resulted in a dramatic reduction in remdesivir dose and schedule leading to lower toxicities, however maintaining a high virological efficacy.

Automated summary

This paper introduces a systematic pathological model learning method for the dynamics of COVID-19, followed by derivative free optimization based multi objective drug rescheduling. By learning from clinical data, the pathological model can predict the immune T cell response and reduce the dose and schedule of remdesivir, reducing toxicity while maintaining virological efficacy.