Batch effects, experimental design and more
1 - Removing batch effects on RNA-seq data
Batch effects are everywhere. The paper proposes a new method for batch effect removal and normalization for RNA-seq data. What I really liked about this paper is that they show how batch effects are hidden in big cohort studies and how careful one should be when analysing them.
2 - Statisticians are not firefighters
A lot of medical research does not follow good practices in experimental design and statistical methodology. The problem is that the statisticians are then called to solve the problems that show up, but only after experiments are performed and data is collected. The comment discuss these problems and how statisticians should be involved in the experiments.
3 - Pitfalls in research design, analysis and reporting
Continuing in the theme from #2, the article below discuss more concrete examples of possible pitfalls in research design, data analysis and reporting that can happen.
4 - Reproducible science in R: package versions
Have you ever tried to run old R code and stumbled in the first lines trying to load the libraries? groundhog is a R library that solves the problem of which package version is the right one when running scripts. By appending the date you used a package in the library call, groundhog is able to load that specific version available from CRAN, Biocondutor or github.
5 - Standards in functional analysis?
Functional analysis is key in transcriptomic and genomic analysis. Despite being everywhere, no standards are used in analysis, making it difficult to even compare results. This paper shows several problems of current functional analysis found in the literature and argues that there is a need for stronger standards in the field.
https://journals.plos.org/ploscompbiol/article?id=10.1371/journal.pcbi.1009935