I want to test effect of a condition across different tissues, and I got some sequencing data from different tissues, and for each tissue, samples are divided into condition A and condition B.

I perform clustering analysis to find variables significantly changed by condition (no matter tissue-specific or global). The heatmap of Z-score matrix shows that the variables' expression pattern is highly tissue-specific but not condition-specific. That is to say, although for different tissues there are differences, samples in same tissue are very similar even if they are in different conditions. In theory, condition A and condition B must make difference. Just in my case, tissue itself has prior impact to condtion. This blocked my further analysis.

So I wonder how to eliminate the effect caused by different tissues, to address the effect of the conditions in data processing.