Data alchemy, from lab to insight: Transforming in vivo experiments into data science gold

Meta-analyses of laboratory-generated data have the potential to improve experimental protocols and offer meaningful understanding of complex biological processes.However, while development of genotype-to-phenotype statistical and predictive studies is growing, in vivogenerated data are generally employed for validation of genotypic models and are not often included as features in the analyses themselves, and rarely used on their own separate from genotypic data.Moreover, the difficulties of normalizing, cleaning, and transforming these data before analysis are multifaceted, in part due to challenges of aggregating sufficient data sets for practical use.Here, we highlight important considerations and best practices when translating data from viral pathogen in vivo studies for use in data science applications, notably statistical analyses, and machine learning (ML) approaches which we use as an illustrative example.Topics covered are applicable when studying multifactorial disease processes (such as viral pathogenicity and transmissibility) independent of the specific data analyses or programming language employed.