February 22, 2021: CADD-Splice manuscript published in Genome Medicine
The new manuscript describes CADD-Splice, the latest extension of CADD to improve its predictions of splicing effects. CADD-Splice is released on this server as CADD v1.6.
July 2, 2020: Fixed off-by-one error in MMSplice annotation for CADD GRCh37-v1.6 and GRCh38-v1.6
We were notified of some inconsistencies between offline scoring without prescored files and online scoring. Tests over many
variants have shown small score differences for 95 out of six million variants.
Most of those were due to a problem with adding MMSplice annotations. Here, the first (in minus strand genes, last) position of any gene was not annotated with MMSplice. This was only an issue for scoring InDel variants, as well as offline scoring of SNVs without prescored files. If you have downloaded the offline scoring scripts and all annotations prior to this fix, please replace the files in the folder `data/annotations/GRCh37-v1.6/mmsplice/` with the updated files for GRCh37 and GRCh38 or re-download all annotations. Please note that you will also have to redownload prescored files for InDels (but not SNVs) if you used any of those files.
Unfortunately, we also noticed a caching problem with the used version of Ensembl VEP. Under rare and uncontrollable circumstances, VEP fails to annotate transcription factor motif scores to GRCh38 variants. This feature has since been fixed in Ensembl VEP, however we cannot update the used VEP version without updating the CADD model. Our next CADD release will update to the most recent Ensembl VEP version and will fix that issue. The frequency of such events is very low. In our test set, we discovered it in 26 out of more than six million variants, with CADD PHRED-score changes of a maximum of 0.4 and only for GRCh38.
We consider these effects very minor and predict that they will not affect most users. We are still sorry for any inconvenience that this may cause. If you have further questions about this issue, please feel free to contact us.
May 26, 2020: Output of multiple annotation lines and SpliceAI tabix index in CADD GRCh37-v1.6
On May 15, we fixed a minor issue with our SpliceAI annotation on GRCh37 that affects chromosome X between 192'989 and 196'609
due to an error in a tabix index. This only affected InDels scored on the webserver and offline scoring without prescored variants.
The annotation release has been updated. If you have downloaded the offline scoring scripts and all annotations prior
to the fix, replace the files in the folder `data/annotations/GRCh37-v1.6/spliceai/` with the updated files from
or re-download all annotations.
On May 25, we noticed that in CADD v1.6 at most one annotation line was returned from pre-scored files. As a result, when retrieving scores for SNVs using our website or offline scoring scripts, for genomic positions overlapping two and more gene annotations, only one annotation line was returned. This has no impact on retrieving the correct CADD score, but might interfere with applications using the returned annotations (also see our note in the info section about using annotation output). Please consider updating your offline scoring scripts of CADD v1.6.
April 11, 2020: Scores and scripts for CADD v1.6 released
With v1.6, CADD integrates two deep learning scores as domain specific-knowledge for splice effect prediction. This release also fixes another issue of the GERP annotation in the GRCh38 model of CADD v1.4/v1.5. More information about the new model can be found in the release notes. The offline installation has been renewed and is now based on Snakemake. Please follow our instructions on the GitHub repository.
February 26, 2019: Scores and scripts for CADD v1.5 released
This release fixes issues in the GERP and ReMap annotations in the GRCh38 model of CADD v1.4. We have further reduced the precision of several annotations in order to decrease the overall file size and updated Ensembl based annotations to release 95. More information about the new model can be found in the release notes. In case you need to update your offline installation, please follow our instructions on the GitHub repository.
October 29, 2018: New CADD manuscript published
The manuscript provides an overview over the changes in CADD from version 1.0 up to 1.4. We further provide some recommendation on how we envision the use of CADD.
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.
August 9, 2018: Browser tracks for UCSC genome browser released, website relaunch, new API and CADD SNV range scoring
We provide Genome Browser Tracks that allow users to visualize CADD versions 1.3 and 1.4 in a UCSC Genome Browser instance for hg19/GRCh37 and hg38/GRCh38. These bigWig tracks were generated using the maximum CADD SNV score for every genome position.
Furthermore, we did a website relaunch that includes several new features for retrieving scores of SNVs (at a single position and now also for coordinate ranges). With the new SNV lookup, we also offer direct links to genome browsers and other resources. Finally, we introduce an API that enables an even simpler retrieval of CADD SNV scores from other applications.
July 4, 2018: Scores and scripts for CADD v1.4 released
The main focus of this release is the support of the genome build GRCh38/hg38, however this version also includes a new CADD model for GRCh37/hg19. We further fixed some minor issues identified in CADD v1.3 with respect to how annotations were interpreted in the model and updated some of the underlying datasets. We included a new splice score (dbscSNV) and measures of genome-wide variant density. The GRCh38 and GRCh37 models are based on the same (or, if these were not available, very similar) annotations. Comprehensive information about the new models can be found in the release notes.
June 27, 2018: GRCh38 release on the horizon
We are actively working on a release of CADD available for both the GRCh37/hg19 and GRCh38/hg38 builds of the human genome. Over the last months, we have collected, curated and created required annotation tracks for GRCh38. We have also updated most of the code base for CADD, making the support of two versions easier for us. Unfortunately, we keep identifying small issues with our release candidates. We hope to release CADD v1.4 very soon...
April 12, 2018: Easy retrieval of individual variants or genomic positions
Based upon frequent requests, we are adding a single variant (SNV) lookup that allows the quick retrieval of CADD scores (and underlying annotations) for a specific genomic position or variant. If interested in a pre-calculated score of a SNV, users no longer need to submit a VCF file through our scoring form or use third-party software (e.g. tabix) to manually retrieve the sites from our pre-scored files.
July 29, 2015: Webserver v1.3 submission rerun for InDels with annotations
We were kindly informed that webform submissions including InDels and requesting results including annotations were missing the InDels for v1.3. The source of the incomplete files was a typo in one of the cluster submission scripts, causing the commands to not be executed. We reran these submissions and ask users to download their files again. If you do not have your download link available, upload your file again and you will be immediately directed to the download page.
July 11, 2015: Scores and scripts for CADD v1.3 released
In CADD v1.3 (3rd developmental/minor release), we fix some minor issues identified with CADD v1.1 and v1.2 related to overlapping gene annotation. This release is still based on the GRCh37/hg19 genome build and we are missing many annotations for the new genome build, not allowing us to release a version for GRCh38 at this point. As for CADD v1.1 and v1.2, many of the results for the validation sets are similar or better than for our last major release (CADD v1.0). We would like to highlight again that there is a measurable reshuffling of variant ranks to the last major release and we see differences in the score ranges that are obtained for certain predicted functional consequences compared to the last major release. Please find further information in the release notes.
For scoring your variants locally with CADD v1.3, we are providing the required annotation tracks (same as for CADD v1.2 with one update), an updated set of primate whole genome alignments and set of scripts. Provided pre-scored variant sets of CADD v1.3 can be used with these scripts to first retrieve pre-scored variants without recalculating their scores. Please check the installation instruction detailed in the README provided with the scripts (132K), before downloading all annotations (206G).
Jan 21, 2015: Scores and scripts for CADD v1.2 released
In CADD v1.2 (2nd developmental/minor release), we fix some minor issues identified with CADD v1.1. CADD v1.2 is still based on the GRCh37/hg19 genome build. As for CADD v1.1, many of the results for the validation sets are similar or better than for our last major release (CADD v1.0). However, we would like to highlight that there is a measurable reshuffling of variant ranks between versions and we see differences in the score ranges that are obtained for certain predicted functional consequences. Please find further information in the release notes.
For scoring your variants locally with CADD v1.2, we are providing the required annotation tracks (same as for CADD v1.1) and a set of scripts. Provided pre-scored variant sets of CADD v1.2 can be used with these scripts to first retrieve pre-scored variants without recalculating their scores. Please check the installation instruction detailed in the README provided with the scripts (132K), before downloading all annotations (206G).
Jan 1, 2015: Scripts and annotations for CADD v1.1 released
If you want to locally score variants without using our webserver, we are now providing the required annotation tracks and a set of scripts for scoring your variants locally with CADD v1.1. You can use the provided pre-scored variant sets to first retrieve pre-scored variants using the provided scripts without recalculating their scores. Please check the installation instruction detailed in the README provided with the scripts (132K), before downloading all annotations (206G). Note that v1.1 has a known bug which will be fixed with the release of CADD v1.2.
Dec 19, 2014: Bug in v1.1 release identified
When multiple SNV variants at the same site are scored, DNA structure predictions for the first variant are considered for all variants in v1.1. This causes the scores distributed in the whole genome SNV files to be off (as three alleles were considered for each position during scoring). This effects the conversion of unscaled to scaled scores. It also causes score differences between the 1000 Genomes, ESP, and ExAC files when compared to the whole genome file. We will fix this in a v1.2 release.
Dec 2, 2014: CADD v1.1 release
In CADD v1.1 (developmental/minor release), we use a slightly extended and updated annotation set and we trained a logistic regression model using GraphLab Create v0.9.1. CADD v1.1 is still based on the GRCh37/hg19 genome build. Even though results reported for many of our previously used validation sets are similar or better with this release, we would like to highlight that there is a measurable reshuffling of variant ranks between versions. We also see differences in the score ranges that are obtained for certain predicted functional consequences. Please find further information in the release notes.
Nov 5, 2014: Scores for ExAC r0.1 variants available for download
ExAC variant scores are available in the download section. These variants were downloaded from ftp://ftp.broadinstitute.org/pub/ExAC_release/release0.1/. Further information about the Exome Aggregation Consortium (ExAC) and the availability of these variants is available at http://exac.broadinstitute.org/
Oct 14, 2014: CADD is running on a new webserver
CADD has been running on a virtual webserver for some time. We have now moved to a new physical webserver with fast local storage. This will allow for faster downloads of the large files that we are providing.
Aug 4, 2014: Set of pre-scored InDels released
If you need to score many variants or score variants on a regular basis, you can download all possible SNVs and this set of InDels and then only score new InDels through the webinterface. In this case, we recommend that you start building a local database of InDels that you previously scored and check against these before running new variants. To initiate a local database of CADD scores, we provide a set of 12.3M InDels for download.
Feb 2, 2014: CADD manuscript is published
Our manuscript describing the method and features of CADD v1.0 was published by Nature Genetics:
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.