Annovar and CADD Annotation

Question

[Ishtiaq Ahmad Khan]

I CADD scored my set of variants called on build 38 by using Annovar. Then I converted my coordinates to hg19 by crossmap and liftover separately and scored them using CADD database and script. The output is quite different. I double checked my coordinate conversion and it is fine. But CADD server and annovar giving different scores. Kindly explain a bit.

[Ishtiaq Ahmad Khan]

Here is a comparison of CADD scores (build 19) for different versions. Why it is so different some time??

CHROM#  Position   REF  Alt  V1.0   V1.1   V1.2   V1.3
1       12854074   C    A    14.29  23     22.4   23.5
1       17301446   G    C    14.3   21.6   19.85  20.8
1       20073719   G    A    17.55  24.9   23.9   25.2
1       22222455   G    T    14.08  11.21  12.46  8.884
1       23966983   G    A    14.2   23.1   22.7   23.3
1       25747156   T    C    13.08  1.574  2.434  9.396
1       45795084   G    A    7.869  11.74  10.95  16.82
1       45974478   G    A    29.2   22.8   22.4   32
1       62579891   G    T    25.5   29.1   28.7   29.4
1       66081833   C    G    13.58  24.9   24     24.7
1       94564483   C    T    24.1   25.3   24.2   24.9
1       109365991  C    T    9.386  21.7   24     23.2
1       148580684  G    A    0.969  0.014  0.014  0.063
1       150483966  G    A    22.3   22.2   22.2   32

Answer 1

[Vivek]

The correlation between CADD scores for the same set of variants over different versions of the CADD release are rather low, especially between version 1.0 and latter versions. You can see the plots from the release notes:

http://cadd.gs.washington.edu/static/ReleaseNotes_CADD_v1.2.pdf

There are differences between the regression models used between the versions as well as updating the feature annotation set used to derive the scores.