findThreshold() returns NA

[Jason Lai]

Hi there,

I am running shazam’s findThreshold() function as part of nf-core/airrflow pipeline. At the findThreshold stage, I noticed that some samples resulted in NA thresholds, as there was no bimodal distribution of the diversity measurement (see attached).

In this scenario, is there a recommended default threshold?

‌

[ssnn]

Hi,

Yes, depending on the distribution of the distances, sometime findThreshold can’t find a threshold. I would use 0.10 or 0.12. We usually also use the distribution of the distances to the nearest sequence in external groups (other subjects) as shown here: https://shazam.readthedocs.io/en/stable/vignettes/DistToNearest-Vignette/#calculating-nearest-neighbor-distances-across-groups-rather-than-within-a-groups In nf-core/airrflow, this will be available in the new release we are finalizing.

‌

‌

[ssnn]

Reopen if needed.