CADD is a tool for scoring the deleteriousness of single nucleotide variants as well as insertion/deletions variants in the human genome (currently supported builds: GRCh37/hg19 and GRCh38/hg38).
Details about CADD, including features in the latest version, the different genome builds and how we envision the use case of CADD are described in our latest manuscript:
Schubach M, Maass T, Nazaretyan L, Roner S, Kircher M.
CADD v1.7: Using protein language models, regulatory CNNs and other nucleotide-level scores to improve genome-wide variant predictions.
Nucleic Acids Res. 2024 Jan 5. doi: 10.1093/nar/gkad989.
PubMed PMID: 38183205.
And and in our previous publications:
Rentzsch P, Schubach M, Shendure J, Kircher M.
CADD-Splice—improving genome-wide variant effect prediction using deep learning-derived splice scores.
Genome Med. 2021 Feb 22. doi: 10.1186/s13073-021-00835-9.
PubMed PMID: 33618777.
Rentzsch P, Witten D, Cooper GM, Shendure J, Kircher M.
CADD: predicting the deleteriousness of variants throughout the human genome.
Nucleic Acids Res. 2018 Oct 29. doi: 10.1093/nar/gky1016.
PubMed PMID: 30371827.
The original manuscript describing the method and its features was published by Nature Genetics in 2014:
Kircher M, Witten DM, Jain P, O'Roak BJ, Cooper GM, Shendure J.
A general framework for estimating the relative pathogenicity of human genetic variants.
Nat Genet. 2014 Feb 2. doi: 10.1038/ng.2892.
PubMed PMID: 24487276.
We provide pre-computed CADD-based scores (C-scores) for all 9 billion possible single nucleotide variants (SNVs) of the reference genome, as well as all SNV and insertions/deletions variants (InDels) from population-wide whole genome variant releases and enable scoring of short InDels on our website.
Please check our website for updates and further information
This section describes how users can setup CADD version 1.7 on their own system. Please note that this requires between 100 GB - 1 TB of disc space and at least 12 GB of RAM.
conda > 24.7.4
We recommend to install conda via miniforge
# For example
curl -L -O "https://github.com/conda-forge/miniforge/releases/latest/download/Miniforge3-$(uname)-$(uname -m).sh"
bash Miniforge3-$(uname)-$(uname -m).shconda install -c conda-forge -c bioconda 'snakemake=8'Note1: The commands are tested with snakemake 8.25.2. This is the minimum snakemake version supported.
unzip CADD.zipinstall.sh to install using a brief installation dialog (see below).This is the easier way of installing CADD, just run:
# if you are located in the US
./install.sh
# if you are located in Europe
./install.sh -bNote that depending on the -b flag, the script will use the German server as download source for annotation files. This may significantly speed up the installation if you are located in Europe.
You first state which parts you want to install (the environments as well as at least one genome build including annotation tracks are neccessary for a quick start) and the script should manage loading and unpacking the neccessary files.
Running CADD depends on four big building blocks (plus the repository containing this README which we assume you already downloaded):
Installing dependencies (when running without apptainer/singularity)
As of this version, dependencies have to be installed via conda and snakemake. This is because we are using two different enviroments for python2 and python3.
snakemake test/input.tsv.gz --software-deployment-method conda --conda-create-envs-only --conda-prefix envs/conda \
--configfile config/config_GRCh38_v1.7.yml --snakefile Snakefile -c 1Please note that we installing both conda environments in the CADD subdirectory envs via --conda-prefix envs. If you do not want this behavior (we do this in order to not install the environments in all active directories you run CADD from), adjust or remove this parameter.
Installing annotations
Both version of CADD (for the different genome builds) rely on a big number of genomic annotations. Depending on which genome build you require you can get them from our website (be careful where you put them as these are really big files and have identical filenames) via:
# if you are located in Europe
cd data/annotations
# for GRCh37 / hg19
wget -c https://kircherlab.bihealth.org/download/CADD/v1.7/GRCh37/GRCh37_v1.7.tar.gz
# for GRCh38 / hg38
wget -c https://kircherlab.bihealth.org/download/CADD/v1.7/GRCh38/GRCh38_v1.7.tar.gz
cd $OLDPWD
# if you are located in the US
cd data/annotations
# for GRCh37 / hg19
wget -c https://krishna.gs.washington.edu/download/CADD/v1.7/GRCh37/GRCh37_v1.7.tar.gz
# for GRCh38 / hg38
wget -c https://krishna.gs.washington.edu/download/CADD/v1.7/GRCh38/GRCh38_v1.7.tar.gz
cd $OLDPWDAs those files are about 300 GB in size, downloads can take long (depending on your internet connection). We recommend to setup the process in the background and using a tool (like wget -c mentioned above) that allows you to continue an interrupted download.
To make sure you downloaded the files correctly, we recommend check md5 hash files from our website (https://kircherlab.bihealth.org/download/CADD/v1.7/GRCh38/MD5SUMs) and checking your downloaded file for completeness (via md5sum GRCh38_v1.7.tar.gz.md5 ; the md5 sums should match the ones we provide at our website).
The annotation files are finally put in the folder data/annotations and unpacked:
cd data/annotations
# for GRCh37 / hg19
tar -zxvf GRCh37_v1.7.tar.gz
# for GRCh38 / hg38
tar -zxvf GRCh38_v1.7.tar.gz
cd $OLDPWDInstalling prescored files
At this point you are ready to go, but if you want a faster version of CADD, you can download the prescored files from our website (see section Downloads for a list of available files). Please note that these files can be very big. The files are (together with their respective tabix indices) put in the folders no_anno or incl_anno depending on the file under data/prescored/${GENOME_BUILD}_${VERSION}/ and will be automatically detected by the CADD.sh script.
You run CADD via the script CADD.sh which technically only requieres an either vcf or vcf.gz input file as last argument. You can further specify the genome build via -g, CADD version via -v (deprecated since version v1.6), request a fully annotated output (-a flag) and specify a seperate output file via -o (else inputfile name .tsv.gz is used). I.e:
./CADD.sh test/input.vcf
./CADD.sh -a -g GRCh37 -o output_inclAnno_GRCh37.tsv.gz test/input.vcfYou can test whether your CADD is set up properly by comparing to the example files in the test directory.
For all command-line inputs please run ./CADD.sh -h.
There are a few options to decrease calculation times of your CADD offline installation:
ESMbatchsize parameter for your local system. This is especially recommented if you have a GPU available. Note that your GPU is automatically detected and set up by the install.sh CADD installation script. You can simply add a line with, e.g., ESMbatchsize: 20 to the the respective CADDv1.7 config file in ./configs/ (e.g., config_GRCh38_v1.7.yml). This will parrallelize calculations of ESM1-v scores on your GPU, which are used for the calculation of CADD scores. This is expected to significantly decrease calculation times. Note that the dafault value for the ESMbatchsize parameter is set to 1 to avoid memory overload. Also note that if changing the EMSbatchsize parameter gives you errors, it is likely that the memory of your GPU is too small. Consider using a smaller number for the parameter in such a case (e.g. ESMbatchsize: 10).Just a CADD-Script version update. CADD scores are the same as with CADD-script v1.7 (CADD scores v1.7). Detailed changes:
docker://visze/cadd-scripts-v1_7:0.1.1VCF2vepVCF.py script fix to extend header. Otherwise regseq will fail using the vcf libraryJust a CADD-Script version update for snakemake 8 compatibility and containerization. CADD scores are the same as with CADD-script v1.7 (CADD scores v1.7). Detailed changes:
Version 1.7 includes some additions in features to v1.6 and we refactured the Snakemake workflow of CADD-scripts including config files. The new models for v1.7 are extended by a protein language model, regulatory effect prediction with a CNN, Zonoomia and Aparent2 scores and an update to the lates gencode 110 version. Because of the refactored workflow with additinal config settings you cannot use it with previous CADD versions. If you are still using those versions, please use this repository for version 1.6 or this repository for v1.5 and v1.4.
Version 1.6 includes some changes in comparison to v1.5. Next to the obvious switch of the pipeline into a Snakemake workflow which became necessary due to the ongoin issues with conda activate, the new models for v1.6 are extended by more specialized annotations for splicing variants, as well as a few minor changes in some other annotations (most prominent: fixed gerp for GRCh38) and changes in consequence categories which make this scripts incompatible with CADD v1.4 and v1.5. If you are still using those version, please use version 1.5 of this repository.
## Copyright
Copyright (c) University of Washington, Hudson-Alpha Institute for
Biotechnology and Berlin Institute of Health at Charite - Universitatsmedizin
Berlin 2013-2024. All rights reserved.
Permission is hereby granted, to all non-commercial users and licensees of CADD
(Combined Annotation Dependent Framework, licensed by the University of
Washington) to obtain copies of this software and associated documentation
files (the "Software"), to use the Software without restriction, including
rights to use, copy, modify, merge, and distribute copies of the Software. The
above copyright notice and this permission notice shall be included in all
copies or substantial portions of the Software.
THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
SOFTWARE.