Hello all, I'd like to propose a rather unorthodox query.
Let's say I have available to me bulk RNA-seq data generated from different labs, on different platforms, on different treatment conditions in the same animal/cell model with all sharing a mutual treatment condition. For example the datasets are as follow: dataset 1 [timepoint 1, timepoint 2], dataset 2 [timepoint 1, timepoint 3], dataset 3 [timepoint 1, timepoint 4]. Raw sequencing data is available and can be re-analysed on the same tools/parameters as needed.
My question is, can I take advantage of the mutual treatment condition (timepoint 1) to try and combine these different datasets to expand the temporal pattern to the gene expression profile? Read counts generated for timepoint 1 would surely be different between datasets (even if they were to be analysed from raw in the same batch), so fold change to timepoint 1 generated from differential analysis will mean little in comparing between different datasets. What I am thinking of is more along the line of normalising the read counts in each dataset to timepoint 1 (using it as an anchor of sort), then use this normalised read count for differential gene expression to get a better sense to differential expression pattern across different datasets.
In scRNA-seq I have come across trajectory/pseudotime analysis which projects the supposed expression of a gene through various conditions ('timepoints'). Can such a thing be done on bulk RNA-seq data? I do realise that such trajectory requires computation of a large number of data points that are available with single-cell data that would not be normally possible with bulk RNA-seq, I just wonder whether it is computationally possible or whether this is completely hogwash.
Or is all of this just simply implausible and I should just be content with the differential expression result that I get from each dataset?