USEARCH manual > UCHIME input order dependency
UCHIME input order dependency
 
See also
UCHIME algorithm
Reproducibility of results

Occasionally, UCHIME de novo produces different results when the input order is changed. This behavior is an unavoidable side-effect of the algorithm design, and does not indicate a bug. The most common reason is a tie in maximum smoothed identity when identifying the top parents. Call the tied parents A1 and A2. The tie is broken arbitrarily (by picking the first in database order). It may be that A1 creates a chimeric 3-way alignment, while A2 does not. This will result in the query sequence being classified as chimeric if A1 is chosen, but not chimeric if A2 is chosen. This is usually because the query is a marginal hit, so the differences with a change in input order is comparable to the change you would see with a small change in the score threshold (-minh option).