When I use your code,dim(ad.count) equal to Null.Why is that?

[wiltbb]

After install,I test your code data('AD_data') ad.count <- AD_data$FullData$X.count dim(ad.count),R returned me Null.By the way,one third of my peanut RNA-seq data samples have batch effect, am I suitable for this R package?

[EvaYiwenWang]

Sorry about that, I changed my data structure. Please try this: data('AD_data') ad.count <- assays(AD_data$FullData)$Count dim(ad.count) For more information, please check here: https://www.bioconductor.org/packages/release/bioc/vignettes/PLSDAbatch/inst/doc/brief_vignette.html.

For your RNA-seq data, could you show me the design of batch and treatment by table(batch, treatment)? I need to check if the design is balanced. In addition, for RNA-seq data, you don't need CLR transformation, you can use other normalisation specific for RNA-seq, then use my PLSDA-batch.

Best, Eva

[wiltbb]

Thank you,Eva,for my RNA-seq data,the yellow part is the RNA-seq data measured by one enzyme, and the green part is the RNA-seq data measured by another enzyme. I want to eliminate the batch effect brought by these two enzymes. Can PLSDA-batch be used for this experimental design?

[wiltbb]

Thank you,Eva,for my RNA-seq data,the yellow part is the RNA-seq data measured by one enzyme, and the green part is the RNA-seq data measured by another enzyme. I want to eliminate the batch effect brought by these two enzymes. Can PLSDA-batch be used for this experimental design?

To summarize, 39 samples used one enzyme, 24 samples used another enzyme, whether the design was balanced?

[wiltbb]

The other problem is that I use other methods, such as combat, removeBatcheffect, etc., to produce negative values, how to deal with these negative values? Will PLSDAbatch have the same problem?

[EvaYiwenWang]

May I know what kind of biological effect you would like to study after batch removal? You cannot get rid of negative values because of the normalisation, like CLR. However, it doesn't matter for downstream analysis like differential expression analysis.

Best, Eva

[wiltbb]

May I know what kind of biological effect you would like to study after batch removal? You cannot get rid of negative values because of the normalisation, like CLR. However, it doesn't matter for downstream analysis like differential expression analysis.

Best, Eva

I will do the differential expression analysis later, how to do it specifically?

[EvaYiwenWang]

There are two options: 1) Univariate methods: You can use simple linear regression (lm()) or a linear mixed model (lmer(), available in the lmerTest R package) if you need to handle random covariates. 2) Multivariate methods: PLSDA (available in the mixOmics R package). You can refer to the following case study: http://mixomics.org/case-studies/splsda-srbct-case-study/. I hope these help.

Best regards, Eva