Skip to main content

Table 1 Method performance on PMD *

From: Better prediction of functional effects for sequence variants

  Method F effect (Eqn. 3) F neutral (Eqn. 3) Q2 (Eqn. 4) MCC (Eqn. 5)
human SNAP2 78.0% ± 0.6 46.3% ± 1.3 68.8% ± 0.7 0.24 ± 0.01
  PolyPhen-2 78.4% ± 0.4 ** 45.1% ± 1.1 ** 68.9% ± 0.5 ** 0.23 ± 0.01 **
  SNAP 74.9% ± 0.5 46.7% ± 1.1 65.8% ± 0.6 0.22 ± 0.01
  SIFT 72.2% ± 0.6 49.0% ± 1.0 63.6% ± 0.6 0.23 ± 0.01
non-human SNAP2 79.9% ± 0.3 45.8% ± 0.8 70.7% ± 0.4 0.26 ± 0.01
  PolyPhen-2 77.1% ± 0.4 44.7% ± 0.8 67.6% ± 0.5 0.22 ± 0.01
  SNAP 77.2% ± 0.3 45.5% ± 0.9 67.9% ± 0.5 0.23 ± 0.01
  SIFT 77.0% ± 0.3 45.8% ± 0.8 67.7% ± 0.4 0.23 ± 0.01
  1. * Data set consisting of 9,657 variants (2,788 neutral, 6,869 effect) from 678 human proteins in the top rows and 42,160 variants (10,850 neutral, 31,310 effect) from 3,383 non-human proteins in the bottom rows. For each measure and species group, significantly best results are highlighted in bold. Measures with no bold highlighting indicate absence of a statistically significant best performer.
  2. ** Values might over-estimate performance for PolyPhen-2 due to overlap between data set used here and one used for training PolyPhen-2.