I am trying to find new markers for senescent cells.

There are multiple datasets i made with fibroblast cells. The difference in the datasets is for example dataset 1 has to measurements. One with data from a young cell, and one from an senescent cell. Some datasets have 2 of young and 1 of senescent cells. My supervisor told me to add all of the datasets together and do rnaseq analysis. Then i should make a pca plot to see what clusters and doe differential expression analysis.

My question is that there is a problem with the clustering, because the cells also are strongly effected by the cell strain. ( they are fibroblasts but from different tissues ). So i came up with the idea to do the pca and de for every dataset separately. To get a list of genes from every dataset and combine them to see if there are overlapping ones.

I have so much fear that this idea is wrong and i will look like a loser if i come up with this idea for my boss. Is this an good idea ?

Typically, it more useful to look at all the batches as a whole.

I will point out that it's very difficult to determine with only 1 replicate per condition within each batch where the variability arises from: within-group, across condition, or across dataset/batch. 

You can analyze this data with a design of ~batch + condition, but it sound very underpowered: not enough samples to detect small or medium sized changes in gene expression.