clintval / chaff / 38431951069

97%

push

github

revert!: fit the copied-damage dispersion per stratum again (#18)

Reverts #16 (8839922).

On heavily damaged duplex libraries, a stratum's single-strand changes often cannot reject a Poisson by the likelihood ratio test, nor can the shape its library's strata share, so the chance model fell back to a Poisson while the stratum's duplex changes still clustered beyond one. Chance then expected more positions with one change than were observed and too few with two or three, and copied damage on two or three molecules scored as real.

BREAKING CHANGE: CDAP, chance_fraction and chance_excess return to their 0.3.0 values on a duplex BAM with single-strand consensus.

28 of 29 new or added lines in 2 files covered. (96.55%)

2322 of 2394 relevant lines covered (96.99%)

383498.01 hits per line