"Solve" the insert size distribution in order to extrapolate the estimate.

[rhpvorderman]

Currently the found insert sizes are displayed. These form a regular statistical distribution. It should be possible to infer a distribution since usually the peak of the distribution is visible. Since one half of the distribution is known, it should be theoretically possible to solve for the parameters and estimate the tail end of the distribution.

[rhpvorderman]

Ioannis suggested I try a few distributions and check with https://en.wikipedia.org/wiki/Kullback%E2%80%93Leibler_divergence

[rhpvorderman]

Apparently the protocol consists of several steps, with first selecting the desired median, then removing smaller and larger inserts in separate steps. This may affect the mapping to a distribution.

[rhpvorderman]

This is infeasible for large insert sizes. It is much better to check this after alignment instead. If the estimate cannot be accurate, it is better not to include it rather than have a potentially misleading graph.