some questions about CMap data

[XiaohanChen97]

Hello!
Recently, I am doing some research using CMap phaseII level5 data. During the research process, I have some questions about the use of control signatrues data which perturbated by DMSO.
Specifically, after dimensionality reduction (using t-SNE) in the expression profile of the control signatrues with the same perturbation concentration and perturbation time, tSNE-1 and tSNE-2（Vectors after dimensionality reduction） could not distinguish the cell lines.On the contrary, after the dimensionality reduction of the expression profile after the use of a certain drug with the same perturbation concentration and perturbation time, tSNE-1 and tSNE-2 can clearly separate the cell line data.
This really puzzles me. Generally speaking, there are obvious differences in the expression profiles of different cancer cell lines. Therefore, the cell line expression profiles of the control signatrues should also have obvious differences. This is not consistent with the conclusion I got from your data.
So I wonder if you could give me some suggestions about what caused this situation, the standardization in the data preprocessing? Or can I get the above conclusion through some data processing procedures?
Finally, I have to say that the CMap database is really a great job!