JMP Genomics - ANOVA vs deseq2

[#1]

Hello there. I am new to both JMP Genomics and RNA-Seq data analysis. It appears that JMP Genomics 10 implements ANOVA as a means to determine differential gene expression data. However, that vast majority of RNA-Seq gene expression studies use deseq2 or similar R packages. Has anyone done a comparison (or know of a published report) comparing the two methods? What are the pros and cons?

Follow up ... Is there any way to implement deseq2 in JMP Genomics?

[#2 · Accepted solution]

Hi @LouisAltamura ,

I am not aware of anyone publishing the difference. I can tell you that ANOVA in JMP G can deal with multiple factors, interactions, nesting and random factors as well as baseline covariates as precursors to differential expression. Something Deseq2 cannot do well or without a lot of work. Also is the interactive nature of the outputs.

They are equivalent in results when looking at similar basic experimental designs. Both use either negative binomial for counts or log2 for transforming data (variance stabilization).  Both can perform Multiple Test Correction methods like Benjamini-Hochberg FDR corrections. There might be some deviation with some options, but they are subtle from what I can tell in regards to the results.

Biggest difference: Coding vs. Point and Click interface

Another related difference: Perceived free (but takes a person time and thus money in order to code) vs. Cost of buying software (but no coding needed)

Since JMP Genomics is based on JMP, one can easily connect to a local copy of R and write JSL to make the connection, invoke Deseq2, and return results back to JMP G/JMP for interactive visualizations. Even provide custom interfaces for non-coders. We have created add-ins for situations like this in Expression/Genetics (QuasR Alignment, Embedding, Genomic Bayesian Regression, and Outcross Linkage Maps) and Predictive Modeling (XGBoost) in JMP Pro/JMP Genomics either connecting to R or Python.