58e83e1a302e901ad4b97a529f15599f25f33e43 jeltje.van.baren Tue Jan 21 11:58:49 2025 -0800 alpha missense description diff --git src/hg/makeDb/trackDb/human/alphaMissense.html src/hg/makeDb/trackDb/human/alphaMissense.html index 462d7df5305..cafe631544c 100644 --- src/hg/makeDb/trackDb/human/alphaMissense.html +++ src/hg/makeDb/trackDb/human/alphaMissense.html @@ -1,172 +1,112 @@ <h2>Description</h2> - -<p> This track collection shows <a href="https://sites.google.com/site/revelgenomics/" -target="_blank">Rare Exome Variant Ensemble Learner</a> (alphaMissense) scores for predicting -the deleteriousness of each nucleotide change in the genome. +<p> +This track shows AlphaMissense predictions for all possible single amino acid substitutions in +the human proteome. </p> - <p> -alphaMissense is an ensemble method for predicting the pathogenicity of missense variants -based on a combination of scores from 13 individual tools: MutPred, FATHMM v2.3, -VEST 3.0, PolyPhen-2, SIFT, PROVEAN, MutationAssessor, MutationTaster, LRT, GERP++, -SiPhy, phyloP, and phastCons. alphaMissense was trained using recently discovered pathogenic -and rare neutral missense variants, excluding those previously used to train its -constituent tools. The alphaMissense score for an individual missense variant can range -from 0 to 1, with higher scores reflecting greater likelihood that the variant is -disease-causing. +AlphaMissense is a deep learning method for predicting the pathogenicity of missense variants +in human proteins. It classifies 32% of all missense variants as likely pathogenic and 57% +as likely benign using a cutoff yielding 90% precision on the ClinVar dataset </p> -<p>Most authors of deleteriousness scores argue against using fixed cutoffs in -diagnostics. But to give an idea of the meaning of the score value, the alphaMissense -authors note: "For example, 75.4% of disease mutations but only 10.9% of -neutral variants (and 12.4% of all ESVs) have a alphaMissense score above 0.5, -corresponding to a sensitivity of 0.754 and specificity of 0.891. Selecting a -more stringent alphaMissense score threshold of 0.75 would result in higher specificity -but lower sensitivity, with 52.1% of disease mutations, 3.3% of neutral -variants, and 4.1% of all ESVs being classified as pathogenic". (Figure S1 of -the reference below) -</p> <h2>Display Conventions and Configuration</h2> -<p> -There are five subtracks for this track: -<ul> -<li> -<p>Four lettered subtracks, one for every nucleotide, showing +<p>There are four lettered subtracks, one for every nucleotide, showing scores for mutation from the reference to that nucleotide. All subtracks show the alphaMissense ensemble score on mouseover. Across the exome, there are three values per position, one for every possible nucleotide mutation. The fourth value, "no mutation", representing the reference allele, e.g. A to A, is always set to zero, "0.0". alphaMissense only takes into account amino acid changes, so a nucleotide change that results in no amino acid change (synonymous) also receives the score "0.0". -</p><p> -In rare cases, two scores are output for the same variant at a -genome position. This happens when there are two transcripts with -different splicing patterns and since some input scores for alphaMissense take into account -the sequence context, the same mutation can get two different scores. In these cases, -only the maximum score is shown in the four per-nucleotide subtracks. The complete set of -scores are shown in the Overlaps track. -</p> - -<li> -<p>One subtrack, Overlaps, shows alternate alphaMissense scores when applicable. -In rare cases (0.05% of genome positions), multiple scores exist with a single variant, -due to multiple, overlapping transcripts. For example, if there are -two transcripts and one covers only half of an exon, then the amino acids -that overlap both transcripts will get two different alphaMissense scores, since some of the underlying -scores (polyPhen for example) take into account the amino acid sequence context and -this context is different depending on the transcript. -For these cases, this subtrack contains at least two -graphical features, for each affected genome position. Each feature is labeled -with the mutation (A, C, T or G). The transcript IDs and resulting score is -shown when hovering over the feature or clicking -it. For the large majority of the genome, this subtrack has no features. -This is because alphaMissense usually outputs only a single score per nucleotide and -most transcript-derived amino acid sequence contexts are identical. -</p> -<p> -Note that in most diagnostic assays, variants are called using WGS -pipelines, not RNA-seq. As a result, variants are originally located on the -genome, not on transcripts, and the choice of transcript is made by -a variant calling software using a heuristic. In addition, clinically, in the -field, some transcripts have been agreed-on as more relevant for a disease, e.g. -because only certain transcripts may be expressed in the relevant tissue. So -the choice of the most relevant transcript, and as such the alphaMissense score, may be -a question of manual curation standards rather than a result of the variant itself. -</p> -</ul> <p> When using this track, zoom in until you can see every basepair at the top of the display. Otherwise, there are several nucleotides per pixel under your mouse cursor and no score will be shown on the mouseover tooltip. </p> <p><b>Track colors</b></p> <p> -This track is colored according to <a target="_blank" href="https://www.sciencedirect.com/science/article/pii/S000292972200461X">Table 2 in Vikas et al</a>. The colors represent the recommended ACMG/AMP score cutoffs. +This track is colored according to the am_class column in the AlphaMissense_$db.tsv file. <table style="text-align: left;"> <thead> <tr> <th>Range</th> <th>Classification</th> </tr> </thead> <tbody> <tr> - <td>≥ .773</td> - <td style="color: rgb(255,0,0);">Pathogenic</td> + <td>≥ .564</td> + <td style="color: rgb(255,0,0);">Likely Pathogenic</td> </tr> <tr> - <td>.772 - .184</td> - <td style="color: rgb(192,192,192);">Neutral</td> + <td>.565 - .340</td> + <td style="color: rgb(192,192,192);">Likely Neutral</td> </tr> <tr> - <td>≤ .183</td> - <td style="color: rgb(80,166,230);">Benign</td> + <td>≤ .340</td> + <td style="color: rgb(80,166,230);">Likely Benign</td> </tr> </tbody> </table> -<p>For hg38, note that the data was converted from the hg19 data using the UCSC -liftOver program, by the alphaMissense authors. This can lead to missing values or -duplicated values. When a hg38 position is annotated with two scores due to the -lifting, the authors removed all the scores for this position. They did the same when -the reference allele has changed from hg19 to hg38. Also, on hg38, the track has -the "lifted" icon to indicate -this. You can double-check if a nucleotide -position is possibly affected by the lifting procedure by activating the track -"Hg19 Mapping" under "Mapping and Sequencing". -</p> <h2>Data access</h2> <p> alphaMissense scores are available at the <a href="https://sites.google.com/site/revelgenomics/" target="_blank"> -alphaMissense website</a>. +alphaMissense cloud storage site</a>. The site provides precomputed alphaMissense scores for all possible human missense variants to facilitate the identification of pathogenic variants among the large number of rare variants discovered in sequencing studies. </p> <p> The alphaMissense data on the UCSC Genome Browser can be explored interactively with the <a href="../cgi-bin/hgTables">Table Browser</a> or the <a href="../cgi-bin/hgIntegrator">Data Integrator</a>. For automated download and analysis, the genome annotation is stored at UCSC in bigWig files that can be downloaded from <a href="http://hgdownload.soe.ucsc.edu/gbdb/$db/alphaMissense/" target="_blank">our download server</a>. The files for this track are called <tt>a.bw, c.bw, g.bw, t.bw</tt>. Individual regions or the whole genome annotation can be obtained using our tool <tt>bigWigToWig</tt> 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 <a href="http://hgdownload.soe.ucsc.edu/downloads.html#utilities_downloads">here</a>. The tools can also be used to obtain features confined to given range, e.g. <br> <br> <tt>bigWigToBedGraph -chrom=chr1 -start=100000 -end=100500 http://hgdownload.soe.ucsc.edu/gbdb/$db/alphaMissense/a.bw stdout</tt> <br> <h2>Methods</h2> <p> Data were converted from the files provided on <a href="https://storage.cloud.google.com/dm_alphamissense" target = "_blank">the alphaMissense Downloads website</a>. As with all other tracks, a full log of all commands used for the conversion is available in our <a target=_blank href="https://github.com/ucscGenomeBrowser/kent/blob/master/src/hg/makeDb/doc/">source repository</a>, for <a target=_blank href="https://raw.githubusercontent.com/ucscGenomeBrowser/kent/master/src/hg/makeDb/doc/hg19.txt">hg19</a> and <a target=_blank href="https://github.com/ucscGenomeBrowser/kent/blob/master/src/hg/makeDb/doc/hg38/alphaMissense.txt">hg38</a>. The release used for each assembly is shown on the track description page. </p> <h2>Credits</h2> <p> -Thanks to the alphaMissense development team for providing precomputed data and fixing duplicated values in the hg38 files. +Thanks to </p> <h2>References</h2> <p> +Cheng J, Novati G, Pan J, Bycroft C, Žemgulytė A, Applebaum T, Pritzel A, Wong LH, +Zielinski M, Sargeant T <em>et al</em>. +<a href="https://www.science.org/doi/abs/10.1126/science.adg7492?url_ver=Z39.88-2003&rfr_id=ori: +rid:crossref.org&rfr_dat=cr_pub%20%200pubmed" target="_blank"> +Accurate proteome-wide missense variant effect prediction with AlphaMissense</a>. +<em>Science</em>. 2023 Sep 22;381(6664):eadg7492. +PMID: <a href="https://www.ncbi.nlm.nih.gov/pubmed/37733863" target="_blank">37733863</a> </p>