The Genome Aggregation Database (gnomAD) - Predicted Constraint Metrics track set contains
metrics of pathogenicity per-gene as predicted for gnomAD v2.1.1 and identifies genes subject to
strong selection against various classes of mutation.
This track includes several subtracks of constraint metrics calculated at gene (canonical
transcript) and transcript level. For more information see the following
The metrics include:
- Observed and expected variant counts per transcript/gene
- Observed/Expected ratio (O/E)
- Z-scores of the observed counts compared to expected
- Probability of loss of function intolerance (pLI), for predicted loss-of-function (pLoF) variation only
Display Conventions and Configuration
There are two "groups" of tracks in this set:
All tracks follow the general configuration settings for bigBed tracks. Mouseover on the
Gene/Transcript Constraint tracks shows the pLI score and the loss of function
observed/expected upper bound fraction (LOEUF), while mouseover on the Regional
Constraint track shows only the missense O/E ratio. Clicking on items in any track brings
up a table of constraint metrics.
- Gene/Transcript LoF Constraint tracks: Predicted constraint metrics at the whole gene
level or whole transcript level for three different types of variation: missense, synonymous,
and predicted loss of function. The Gene Constraint track displays metrics for a canonical
transcript per gene defined as the longest isoform. The Transcript Constraint track displays
metrics for all transcript isoforms. Items on both tracks are shaded according to the pLI score,
with outlier items shaded in grey.
- Gene/Transcript Missense Constraint tracks: The missense constraint tracks are built
similarly to the LoF constraint tracks, however the items displayed are based on
missense Z scores.
All items are colored black, and individual Z scores can be seen on mouseover.
Clicking the grey box to the left of the track, or right-clicking and choosing the Configure option,
brings up the interface for filtering items based on their pLI score, or labeling the items
based on their Ensembl identifier and/or Gene Name.
Please see the gnomAD browser help page and FAQ for further explanation of the topics below.
Observed and Expected Variant Counts
Observed count: The number of unique single-nucleotide variants in each transcript/gene
with 123 or fewer alternative alleles (MAF < 0.1%).
Expected count: A depth-corrected probability prediction model that takes into account
sequence context, coverage, and methylation was used to predict expected
variant counts. For more information please see Lek et al., 2016.
Variants found in exons with a median depth < 1 were removed from both counts.
The O/E constraint score is the ratio of the observed/expected variants in that gene. Each item in
this track shows the O/E ratio for three different types of variation: missense, synonymous, and
loss-of-function. The O/E ratio is a continuous measurement of how tolerant a gene or
transcript is to a certain class of variation. When a gene has a low O/E value, it is under stronger
selection for that class of variation than a gene with a higher O/E value. Because Counts depend on
gene size and sample size, the precision of the values varies a lot from one gene to the next.
Therefore, the 90% confidence interval (CI) is also displayed along with the O/E ratio to better
assist interpretation of the scores.
When evaluating how constrained a gene is, it is essential to consider the CI when using O/E. In
research and clinical interpretation of Mendelian cases, pLI > 0.9 has been widely used for
filtering. Accordingly, the Gnomad team suggests using the upper bound of the O/E confidence interval
LOEUF < 0.35 as a threshold if needed.
Please see the Methods section below for more information about how the scores were calculated.
pLI and Z-scores
The pLI and Z-scores of the deviation of observed variant counts relative to the expected number
are intended to measure how constrained or intolerant a gene or transcript is to a specific type of
variation. Genes or transcripts that are particularly depleted of a specific class of variation
(as observed in the gnomAD data set) are considered intolerant of that specific type of variation.
Z-scores are available for the missense and synonynmous categories and pLI scores are available for
the loss-of-function variation.
Missense and Synonymous: Positive Z-scores indicate more constraint (fewer observed
variants than expected), and negative scores indicate less constraint (more observed variants than
expected). A greater Z-score indicates more intolerance to the class of variation. Z-scores
were generated by a sequence-context-based mutational model that predicted the number of expected
rare (< 1% MAF) variants per transcript. The square root of the chi-squared value of the
deviation of observed counts from expected counts was multiplied by -1 if the observed count was
greater than the expected and vice versa. For the synonymous score, each Z-score was corrected by
dividing by the standard deviation of all synonymous Z-scores between -5 and 5. For the missense
scores, a mirrored distribution of all Z-scores between -5 and 0 was created, and then all missense
Z-scores were corrected by dividing by the standard deviation of the Z-score of the mirror
Loss-of-function: pLI closer to 1 indicates that the gene or transcript cannot tolerate
protein truncating variation (nonsense, splice acceptor and splice donor variation). The gnomAD
team recommends transcripts with a pLI >= 0.9 for the set of transcripts extremely intolerant
to truncating variants. pLI is based on the idea that transcripts can be classified into three
An expectation-maximization algorithm was then used to assign a probability of belonging in each
class to each gene or transcript. pLI is the probability of belonging in the haploinsufficient class.
- null: heterozygous or homozygous protein truncating variation is completely tolerated
- recessive: heterozygous variants are tolerated but homozygous variants are not
- haploinsufficient: heterozygous variants are not tolerated
Please see Samocha et al., 2014 and Lek et al., 2016 for further discussion of these metrics.
Transcripts from GENCODE v40 were filtered according to the following criteria:
- Must have methionine at start of coding sequence
- Must have stop codon at end of coding sequence
- Must be divisible by 3
- Must have at least one observed variant when removing exons with median depth < 1
- Must have reasonable number of missense and synonymous variants as determined by a Z-score cutoff
UCSC Track Methods
Gene and Transcript Constraint tracks
Per gene and per transcript data were downloaded from the gnomAD Google Storage bucket:
These data were then joined to the Gencode v40 set of genes/transcripts available at the UCSC
Genome Browser and then transformed into a bigBed 12+5. For the full list of commands used to
make this track please see the
The raw data can be explored interactively with the Table Browser, or
the Data Integrator. For automated access, this track, like all
others, is available via our API. However, for bulk
processing, it is recommended to download the dataset. The genome annotation is stored in a bigBed
file that can be downloaded from the
download server. The exact
filenames can be found in the track configuration file. Annotations can be converted to ASCII text
by our tool
bigBedToBed which can be compiled from the source code or downloaded as
a precompiled binary for your system. Instructions for downloading source code and binaries can be
found here. The tool
can also be used to obtain only features within a given range, for example:
bigBedToBed http://hgdownload.soe.ucsc.edu/gbdb/hg38/gnomAD/pLI/pliByTranscript.bb -chrom=chr6 -start=0 -end=1000000 stdout
Please refer to our
mailing list archives
for questions and example queries, or our
Data Access FAQ
for more information.
More information about using and understanding the gnomAD data can be found in the
gnomAD FAQ site.
Thanks to the Genome Aggregation
Database Consortium for making these data available. The data are released under the ODC Open Database License
(ODbL) as described here.
Lek M, Karczewski KJ, Minikel EV, Samocha KE, Banks E, Fennell T, O'Donnell-Luria AH, Ware JS, Hill
AJ, Cummings BB et al.
Analysis of protein-coding genetic variation in 60,706 humans.
Nature. 2016 Aug 18;536(7616):285-91.
PMID: 27535533; PMC: PMC5018207
Karczewski KJ, Francioli LC, Tiao G, Cummings BB, Alföldi J, Wang Q, Collins RL, Laricchia KM,
Ganna A, Birnbaum DP et al.
The mutational constraint spectrum quantified from variation in 141,456 humans.
Nature. 2020 May;581(7809):434-443.
PMID: 32461654; PMC: PMC7334197
Collins RL, Brand H, Karczewski KJ, Zhao X, Alföldi J, Francioli LC, Khera AV, Lowther C,
Gauthier LD, Wang H et al.
A structural variation reference for medical and population genetics.
Nature. 2020 May;581(7809):444-451.
PMID: 32461652; PMC: PMC7334194
Cummings BB, Karczewski KJ, Kosmicki JA, Seaby EG, Watts NA, Singer-Berk M, Mudge JM, Karjalainen J,
Satterstrom FK, O'Donnell-Luria AH et al.
Transcript expression-aware annotation improves rare variant interpretation.
Nature. 2020 May;581(7809):452-458.
PMID: 32461655; PMC: PMC7334198