一个基因有多个转录本，而且组织具有组织表达特异性，即不同的组织会表达不同的转录本。而昂飞芯片在设计的时候考虑了这个问题，因此大部分探针是设计在通用区域，有部分基因设计了多个转录本。所以一般情况，你分析时，按表达量高的来分析好了，除非你对不同转录本有不同需求。

before you do any of that see if you can associate your probes to transcripts (ENST or otherwise) instead of genes. You might find some of these changes are limited to one isoform, which you would mask with the averaging.

this option is not on your list, but I would check to see if the probes impart information about multiple transcripts from the same gene. I don't know of a way to do this programmatically, but if you have a list of top hit probes, you could do it for those, then revisit even non-significant probes from within the same gene. If there are any. If you are calculating the fold change, then you want to show the biggest difference because this is likely a list of top hits or something, so I would go for option 2 or 8.