There will likely be significant batch effects. I would analyze each set separately to get a higher power (which will be the case when the variability between the sets is large and won't be compensated by the reduction of standard errors due to the larger sample size).

You might consider pooling the p-values according to Fisher's method, if you need a single p-value per gene.

Whether to merge data sets or not depends on the biological context behind the analysis.

Merge the dataset when analysing (a) differential expression; (b) same tissue; (c) ease batch analysis to remove unwanted variation; (d) normalization of datasets; and (e) visualize the clustering of samples.

A separate analysis of datasets (a) Meta-analysis to find common mRNAs/differentially expressed genes; (b) check the consistency of the differential expression; and (c) understand the biological pathways enriched in each dataset

I would suggest you select as per the objective or do both and compare the results using the respective visualization.

I hope this helps. Regards,

Hello Sanjukta, I would recommend not merging three datasets because each of them might contain different patient samples collected under different conditions.

Thank you for your inputs. Samvedna Singh Tikshana Yadav Jochen Wilhelm

Good Luck!

It is not recommended due to experimental variation as well individual among other including machinery and preprocessing. However you can merge and analyze through clustering algorithm to see if they are clustered near or far to get idea about the variation exist in each of dataset.

@shahzaib khoso

Can you mention some open clustering algorithm to check variations, as you mentioned.