Missing data affects how scientists count biological sources
Researchers found that removing specific references can change how many sources explain a distance profile.
Scientists study how many sources explain a distance profile. This process is called phylogenetic distance deconvolution. It involves placing weighted sources on a reference tree.
The study looks at the minimum number of strictly interior sources. It also examines the cost of having incomplete references. An exact algorithm exists for a positive-rational reduced tree with a free positive mean pendant.
Removing references can only lower the minimum source count. It can never raise it. This was held without exception in 23,545 draws. Which references are missing matters more than how many are missing.
Two adversarially chosen references suffice at the median to destroy uniqueness. This number does not grow with panel size. Random loss scales linearly. Discarding 7,931 of 10,575 references from a real 16S panel at random costs 1.8% of the tolerance.
Discarding comparably many as whole clades costs 90%. This tolerance needs neither ground truth nor an error model. Published distance estimates exceed it in 296 of 296 cases.
The phenomenon is unobservable through a point estimate of the count. Methods for this problem should report the tolerance at which an explanation of a given size exists.
Key facts
- Removing references can only lower the minimum source count and never raise it.
- Discarding 7,931 of 10,575 references from a real 16S panel at random costs 1.8% of the tolerance.
- Two adversarially chosen references can destroy the uniqueness of the minimum explanation.