This paper studied proteomic and metabolic characterization of COVID-19 patient sera to develop early detection tools to help predict early on which patients are most likely to progress to severe and life-threatening illness. Guo and colleagues analyzed hundreds of molecular changes in blood samples collected from 53 healthy participants and 46 participants with COVID-19, including 21 with severe disease. The researchers then trained machine learning to discern patterns or molecular signatures to determine if it could learn to distinguish between mild and severe COVID-19 based on molecular data alone.  

The results showed that the machine learning model could differentiate mild and serve COVID-19 based on 22 proteins and 7 metabolites in a training cohort of 18 non-severe and 13 severe patients with an overall accuracy of 93.5% in the training set. In further prospective validation tests, the data from such tests might be useful for prioritizing therapeutic strategies for the severe patients.  

However, there are limitations of this study. First, the gap in age between severe and non-severe patients is 12 years, which could have affected the precision of data interpretation. Moreover, the severe patients also exhibit slightly higher BMI and a higher proportion of some comorbidities. Finally, the sample size is rather small. Future studies of sera from more time points are required for temporal analysis.