I have just started to analyze some variants-of-interests in NK-type of cells. I gotten the plots out but I was puzzled when some of the overlaid “REF+ALT” tracks only have the ‘ALT’ predictions/values plotted.
I have pasted a plot from AlphaGenome, overlooking a positional variant across multiple predicted omics. In the plot, the first few RNA-seq tracks are all ‘red’ and missing out on the REF (gray) values. Just wondering if this is a ‘bug’ or simply due to my misunderstanding of the output?
I think the ALT+REFs tracks are all being plotted, but the REFs either have no/very little effect from this zoom level. The legends on the right I believe are only drawn once for each output type, which might be adding to the confusion.
Thanks for your reply. I suspect the context of the input sequence was too short for the model to confidently give an output for the REFs and, hence, the REFs was ‘blanks’. Howeer, going along my line of reasoning, the REFs should be in the pretraining phase much more often. Then shouldnt the ALTs be ‘blanks’ more often than the REFs?
