If the 5' read count is very high compared to the EXPR_CONTROL read count, the 3'/5'
Imbalance value is not a reliable predictor. With high 5' read counts, the potential for
false negatives (missing a real fusion) increases.
This example plot shows a false negative in sample 32. The green line tracks the 5' expression
as 5' read counts divided by EXPR_CONTROL read counts. Note the spike of high 5'
expression at the false negative for sample 21.
We have not seen a high 5' expression to be associated with a false positive.
