What does a question mark ("?") mean in Picard metrics files when I expect a number (integer, float, etc.)

Question

Does anyone know what conditions must be true for Picard to put a ? where I expect a number?

Here is an example from the output of the tool CollectWgsMetrics:

❯ cat sample.wgs_metrics.txt | head -n8 | tail -n2 | verticalize | grep FOLD
FOLD_80_BASE_PENALTY    1.33476
FOLD_90_BASE_PENALTY    2.288159
FOLD_95_BASE_PENALTY    ?

Code searching (link) didn't reveal anything, and googling isn't returning anything obvious either.

---

[Verticalize is a little CLI tool I've forever been using by Pierre Lindenbaum: verticalize. Easy to compile and I toss it in my ~/bin every time I get a new computer.]

Answer 1

Picard (actually HtsJdk) uses a formatting class (via the MetricsFile class) that emits "?" for NaNs (and Infs).

In your case, there seems to be a division-by-zero error during the calculation of the metric (likely due to zero coverage at some small percentage of bases in your input). This resulting Nan is emitted as a "?".