CoChat
Sign Up
Cited by 12Genomics and Rare DiseasesGenetic Associations and EpidemiologyCancer Genomics and Diagnostics

The DOI matches its Crossref record; 5 of the 5 references it cites were found in a registry or an index.

DeMAG predicts the effects of variants in clinically actionable genes by integrating structural and evolutionary epistatic features

Open access
Luppino, Federica, Adzhubei, Ivan A., Cassa, Christopher A., Toth-Petroczy, Agnes·2023·Nature Communications
DOI 10.1038/s41467-023-37661-zPMC PMC10115847

Published version of a preprint

Open this paper in CoChat — free accountOpen PDF

Abstract

Despite the increasing use of genomic sequencing in clinical practice, the interpretation of rare genetic variants remains challenging even in well-studied disease genes, resulting in many patients with Variants of Uncertain Significance (VUSs). Computational Variant Effect Predictors (VEPs) provide valuable evidence in variant assessment, but they are prone to misclassifying benign variants, contributing to false positives. Here, we develop Deciphering Mutations in Actionable Genes (DeMAG), a supervised classifier for missense variants trained using extensive diagnostic data available in 59 actionable disease genes (American College of Medical Genetics and Genomics Secondary Findings v2.0, ACMG SF v2.0). DeMAG improves performance over existing VEPs by reaching balanced specificity (82%) and sensitivity (94%) on clinical data, and includes a novel epistatic feature, the ‘ partners score’, which leverages evolutionary and structural partnerships of residues. The ‘partners score’ provides a general framework for modeling epistatic interactions, integrating both clinical and functional information. We provide our tool and predictions for all missense variants in 316 clinically actionable disease genes (demag.org) to facilitate the interpretation of variants and improve clinical decision-making.

Overview

The claims and statistics of this paper haven’t been checked yet.

Run the check — free, no account needed. It reads the open-access text and checks the sources it cites.

A few free checks a day · papers already verified open instantly · results are public

Integrity signals

Registry

Crossref · 10.1038/s41467-023-37661-z

Cites

5 of 5 references verified0 citation errors · 0 not checkable

Statistics

Figures

Fig. 1 Overview of DeMAG (Deciphering Mutations in Actionable Genes). First, we assembled a training set from 59 actionable genes; pathogenic variants were collected from clinical databases such as ClinVar and Human Gene Mutation Database (HGMD40). The benign class includes variants from different sources including clinical, population, and non-human primate variants. The training set consisted of 6713 (60%) pathogenic and 4512 (40%) benign variants. Next, we annotated each variant with features, including EVmutation18, IUPred2A75, and AlphaFold240 confidence score (predicted Local Distance Difference Test, pLDDT). We designed a novel feature, the ‘partners score’, that captures the epistatic effects both in the sequence and in the 3D space of the protein. It relies on the observation that evolutionary coupled and spatially close residues are enriched in the same phenotypic effect. Next, we trained a classification tree-based gradient-boosting model using 13 selected features. The model was validated with 3 different types of data: (1) clinical testing data from ClinVar, (2) benign variants from the Estonian Biobank, and (3) deep mutational scanning data for four genes (BRCA1, TP53, MSH2, and PTEN). Finally, we provide predictions for all variants in the 59 actionable genes and extra 257 ClinVar genes and a web application (demag.org). Created with BioRender.com.
Fig. 1 Overview of DeMAG (Deciphering Mutations in Actionable Genes). First, we assembled a training set from 59 actionable genes; pathogenic variants were collected from clinical databases such as ClinVar and Human Gene Mutation Database (HGMD40). The benign class includes variants from different sources including clinical, population, and non-human primate variants. The training set consisted of 6713 (60%) pathogenic and 4512 (40%) benign variants. Next, we annotated each variant with features, including EVmutation18, IUPred2A75, and AlphaFold240 confidence score (predicted Local Distance Difference Test, pLDDT). We designed a novel feature, the ‘partners score’, that captures the epistatic effects both in the sequence and in the 3D space of the protein. It relies on the observation that evolutionary coupled and spatially close residues are enriched in the same phenotypic effect. Next, we trained a classification tree-based gradient-boosting model using 13 selected features. The model was validated with 3 different types of data: (1) clinical testing data from ClinVar, (2) benign variants from the Estonian Biobank, and (3) deep mutational scanning data for four genes (BRCA1, TP53, MSH2, and PTEN). Finally, we provide predictions for all variants in the 59 actionable genes and extra 257 ClinVar genes and a web application (demag.org). Created with BioRender.com.CC BY · PubMed Central
Fig. 2 The partners score – integrating evolutionary and structural information to inform variant assessment. a Co-evolving residue pairs in the protein sequence and spatially adjacent residues (<11 Å) in the protein 3D structure. Residues are colored according to the phenotypic effect of associated variants within the DeMAG training set, e.g., red for positions with only pathogenic variants (see legend at left). The partners score informs residues positions that do not have any associated variants but are either co-evolving or spatially close to residue positions that have variants with known phenotypic effect (i.e., annotated in the training set). b A point scale quantifies the connection between residue positions e.g., benign positions are assigned −1, pathogenic 1, and mixed or not annotated 0. Next, we defined the residue score as the sum of partners points per residue e.g., residue glutamic acid (E) is connected with two pathogenic variants thus its residue score is 2. The partners score is derived via a mixture discriminant analysis model. First, the residue score distribution (histograms) for the benign and pathogenic residue positions is estimated with a gaussian mixture model (density lines). c The posterior probability of belonging both to the pathogenic and benign class is computed given the residue score and the prior probability (i.e., frequency of the two classes). The partners score is defined as the posterior probability of pathogenicity. The partners score feature is available for 74% of ACMG SF v2.0 residues positions, while only 13% of positions have annotations in the training set. Created with BioRender.com.
Fig. 2 The partners score – integrating evolutionary and structural information to inform variant assessment. a Co-evolving residue pairs in the protein sequence and spatially adjacent residues (<11 Å) in the protein 3D structure. Residues are colored according to the phenotypic effect of associated variants within the DeMAG training set, e.g., red for positions with only pathogenic variants (see legend at left). The partners score informs residues positions that do not have any associated variants but are either co-evolving or spatially close to residue positions that have variants with known phenotypic effect (i.e., annotated in the training set). b A point scale quantifies the connection between residue positions e.g., benign positions are assigned −1, pathogenic 1, and mixed or not annotated 0. Next, we defined the residue score as the sum of partners points per residue e.g., residue glutamic acid (E) is connected with two pathogenic variants thus its residue score is 2. The partners score is derived via a mixture discriminant analysis model. First, the residue score distribution (histograms) for the benign and pathogenic residue positions is estimated with a gaussian mixture model (density lines). c The posterior probability of belonging both to the pathogenic and benign class is computed given the residue score and the prior probability (i.e., frequency of the two classes). The partners score is defined as the posterior probability of pathogenicity. The partners score feature is available for 74% of ACMG SF v2.0 residues positions, while only 13% of positions have annotations in the training set. Created with BioRender.com.CC BY · PubMed Central
Fig. 3 The partners score annotates residue positions lacking prior functional or clinical assessments. a On the top left, co-evolving residue positions are shown for the DNA mismatch repair protein MSH6. The innermost circle indicates residues that are co-evolving with at least one other residue whose phenotypic effect is known (i.e., annotated in the training set). Among such residues, 8 are pathogenic, 7 benign, and 255 lack prior assessment. The outer circle shows the partners score resulting from co-evolutionary partnership of residues, which gives a score to 255 residue positions without known annotations. The outermost circle indicates the Pfam domains of MSH6. On the right, AlphaFold 3D model of the protein is shown. Residue positions are colored based on the partners score derived from spatially close partnerships of residues, which inform 750 previously unannotated residue positions (55% of the protein length). Below, the same representation is shown for the cellular tumor antigen P53 protein. The circle plot shows correlation between partners score and domain annotation: the residues that belong to the DNA-binding domain have high partners scores while in the low-complexity region have low partners scores (Supplementary Fig. 7). In total 208 (53%) previously unlabeled residue positions are now annotated with partner scores. b A structural example, where MSH6 ATP binding site residues, have been previously shown50 to be critical for DNA mismatch repair (MMR) function, have a high partners score.
Fig. 3 The partners score annotates residue positions lacking prior functional or clinical assessments. a On the top left, co-evolving residue positions are shown for the DNA mismatch repair protein MSH6. The innermost circle indicates residues that are co-evolving with at least one other residue whose phenotypic effect is known (i.e., annotated in the training set). Among such residues, 8 are pathogenic, 7 benign, and 255 lack prior assessment. The outer circle shows the partners score resulting from co-evolutionary partnership of residues, which gives a score to 255 residue positions without known annotations. The outermost circle indicates the Pfam domains of MSH6. On the right, AlphaFold 3D model of the protein is shown. Residue positions are colored based on the partners score derived from spatially close partnerships of residues, which inform 750 previously unannotated residue positions (55% of the protein length). Below, the same representation is shown for the cellular tumor antigen P53 protein. The circle plot shows correlation between partners score and domain annotation: the residues that belong to the DNA-binding domain have high partners scores while in the low-complexity region have low partners scores (Supplementary Fig. 7). In total 208 (53%) previously unlabeled residue positions are now annotated with partner scores. b A structural example, where MSH6 ATP binding site residues, have been previously shown50 to be critical for DNA mismatch repair (MMR) function, have a high partners score.CC BY · PubMed Central
Fig. 4 Epistatic and structural features improve DeMAG performance, both within and across genes in the training set. a Performance within genes: for most genes, variants are classified with high ROC-AUC > 70%. The high performance is also maintained for the classification of pathogenic (sensitivity >70%) and benign (specificity >70%) variants in most genes. b The feature importance of the 13 DeMAG features averaged across n = 4 cross validation (CV) folds (see Supplementary Table 2 for features description). The whiskers of the boxplots range correspond to ±1.5 times the IQR (inter quantile range). The lower (upper) bound of the box of the boxplot corresponds to the 25th (75th) percentile. The center is the median. The partners score feature stands out from the others with a median importance of ~60%. c Performance metrics for DeMAG. The metrics represent mean values obtained by 4-fold CV with standard deviation in parentheses. Notably, DeMAG shows balanced performance with 87% sensitivity and 85% specificity as well as 90% positive predicted value and 80% negative predictive value. Removing epistatic and structural features provides consistently lower performance than the complete model.
Fig. 4 Epistatic and structural features improve DeMAG performance, both within and across genes in the training set. a Performance within genes: for most genes, variants are classified with high ROC-AUC > 70%. The high performance is also maintained for the classification of pathogenic (sensitivity >70%) and benign (specificity >70%) variants in most genes. b The feature importance of the 13 DeMAG features averaged across n = 4 cross validation (CV) folds (see Supplementary Table 2 for features description). The whiskers of the boxplots range correspond to ±1.5 times the IQR (inter quantile range). The lower (upper) bound of the box of the boxplot corresponds to the 25th (75th) percentile. The center is the median. The partners score feature stands out from the others with a median importance of ~60%. c Performance metrics for DeMAG. The metrics represent mean values obtained by 4-fold CV with standard deviation in parentheses. Notably, DeMAG shows balanced performance with 87% sensitivity and 85% specificity as well as 90% positive predicted value and 80% negative predictive value. Removing epistatic and structural features provides consistently lower performance than the complete model.CC BY · PubMed Central
Fig. 5 DeMAG outperforms other VEPs on clinical variants. Pairwise comparison of the performance of DeMAG and other VEPs on 1285 high-quality clinical variants from the ClinVar testing set (852 pathogenic and 433 benign variants), according to different performance metrics (specificity, sensitivity, Matthews Correlation Coefficient (MCC), ROC-AUC and accuracy). The left panel shows specificity and variants coverage of VEPs (squares) (e.g., SIFT has 95% coverage, see Table 1 for exact values). The green horizontal line (y = 1) indicates that DeMAG has 100% coverage, it has prediction for all variants. DeMAG’s high specificity is on par with EVE on its 73% variants coverage (thin black bordersquare). For EVE there are 2 comparisons: the first (bars without black border) indicate the performance when uncertain class variants are excluded and the bars with black borders include the uncertain variants as well. We assigned their class as if EVE was a random classifier (Table 1). While all tools reach almost perfect ROC-AUC and sensitivity, DeMAG has the most balanced performance, namely the highest MCC (top right panel). For confidence intervals calculated on 1000 bootstrap samples refer to Supplementary Fig. 14.
Fig. 5 DeMAG outperforms other VEPs on clinical variants. Pairwise comparison of the performance of DeMAG and other VEPs on 1285 high-quality clinical variants from the ClinVar testing set (852 pathogenic and 433 benign variants), according to different performance metrics (specificity, sensitivity, Matthews Correlation Coefficient (MCC), ROC-AUC and accuracy). The left panel shows specificity and variants coverage of VEPs (squares) (e.g., SIFT has 95% coverage, see Table 1 for exact values). The green horizontal line (y = 1) indicates that DeMAG has 100% coverage, it has prediction for all variants. DeMAG’s high specificity is on par with EVE on its 73% variants coverage (thin black bordersquare). For EVE there are 2 comparisons: the first (bars without black border) indicate the performance when uncertain class variants are excluded and the bars with black borders include the uncertain variants as well. We assigned their class as if EVE was a random classifier (Table 1). While all tools reach almost perfect ROC-AUC and sensitivity, DeMAG has the most balanced performance, namely the highest MCC (top right panel). For confidence intervals calculated on 1000 bootstrap samples refer to Supplementary Fig. 14.CC BY · PubMed Central
Fig. 6 Comparison of VEPs on population data. Misclassification rate of benign common variants within the Estonian Biobank. The test set constitutes of 94 variants. On the y axis the misclassification rate and on the second y axis the VEP’s variants coverage, identified by the colored squares. The green horizontal line (y = 1) shows that DeMAG has 100% coverage, as wells as REVEL, M-CAP and ClinPred. Among those tools DeMAG reaches the lowest misclassification rate. VARITY reaches the lowest misclassification rate but covers only 80% of variants. The comparison with EVE has 2 error bars: the first (black border and transparent color) indicates the performance when uncertain class variants are predicted as if EVE was a random classifier, the inner (and shorter) bar indicates EVE’s performance excluding uncertain variants. While EVE has lower misclassification rate than DeMAG, it only gives certain predictions for 65% of variants. Although VARITY and EVE classify these 94 variants with the lowest misclassification rate, DeMAG has the lowest misclassification rate among VEPs with 100% predictions. For confidence intervals calculated on 1000 bootstrap samples refer to Supplementary Fig. 15.
Fig. 6 Comparison of VEPs on population data. Misclassification rate of benign common variants within the Estonian Biobank. The test set constitutes of 94 variants. On the y axis the misclassification rate and on the second y axis the VEP’s variants coverage, identified by the colored squares. The green horizontal line (y = 1) shows that DeMAG has 100% coverage, as wells as REVEL, M-CAP and ClinPred. Among those tools DeMAG reaches the lowest misclassification rate. VARITY reaches the lowest misclassification rate but covers only 80% of variants. The comparison with EVE has 2 error bars: the first (black border and transparent color) indicates the performance when uncertain class variants are predicted as if EVE was a random classifier, the inner (and shorter) bar indicates EVE’s performance excluding uncertain variants. While EVE has lower misclassification rate than DeMAG, it only gives certain predictions for 65% of variants. Although VARITY and EVE classify these 94 variants with the lowest misclassification rate, DeMAG has the lowest misclassification rate among VEPs with 100% predictions. For confidence intervals calculated on 1000 bootstrap samples refer to Supplementary Fig. 15.CC BY · PubMed Central

Verification

Claims the paper makes

The claims and statistics of this paper haven’t been checked yet.

Run the check — free, no account needed. It reads the open-access text and checks the sources it cites.

A few free checks a day · papers already verified open instantly · results are public

Self-check

Percentages, sums and reported differences re-computed from the paper’s own numbers.

The self-check needs the paper’s full text.

Citations

Cites

Research works printed but not linked: 67

  1. Miller, D. T. et al. ACMG SF v3.0 list for reporting of secondary findings in clinical exome and genome sequencing: a policy statement of the American College of Medical Genetics and Genomics (ACMG). Genet. Med. 23, 1381–1390 (2021). · DOI
  2. Kalia, S. S. et al. Recommendations for reporting of secondary findings in clinical exome and genome sequencing, 2016 update (ACMG SF v2.0): a policy statement of the American College of Medical Genetics and Genomics. Genet. Med. 19, 249–255 (2017). · DOI
  3. Green, R. C. et al. ACMG recommendations for reporting of incidental findings in clinical exome and genome sequencing. Genet. Med. 15, 565–574 (2013). · DOI
  4. Landrum, M. J. et al. ClinVar: public archive of relationships among sequence variation and human phenotype. Nucleic Acids Res. 42, D980–D985 (2014). · DOI · Paper page
  5. Makhnoon, S., Shirts, B. H. & Bowen, D. J. Patients’ perspectives of variants of uncertain significance and strategies for uncertainty management. J. Genet. Couns. 28, 313–325 (2019). · DOI
  6. Murray, M. L., Cerrato, F., Bennett, R. L. & Jarvik, G. P. Follow-up of carriers of BRCA1 and BRCA2 variants of unknown significance: Variant reclassification and surgical decisions. Genet. Med. 13, 998–1005 (2011). · DOI
  7. Ong, M.-S. & Mandl, K. D. National expenditure for false-positive mammograms and breast cancer overdiagnoses estimated at $4 billion a year. Health Aff. 34, 576–583 (2015). · DOI
  8. Adzhubei, I. A. et al. A method and server for predicting damaging missense mutations. Nat. Methods 7, 248–249 (2010). · DOI · Paper page
  9. Carter, H., Douville, C., Stenson, P. D., Cooper, D. N. & Karchin, R. Identifying Mendelian disease genes with the Variant Effect Scoring Tool. BMC Genomics 14, S3 (2013). · DOI
  10. Jagadeesh, K. A. et al. M-CAP eliminates a majority of variants of uncertain significance in clinical exomes at high sensitivity. Nat. Genet. 48, 1581–1586 (2016). · DOI
  11. Ioannidis, N. M. et al. ARTICLE REVEL: an ensemble method for predicting the pathogenicity of rare missense variants. Am. J. Hum. Genet. 99, 877–885 (2016). · DOI · Paper page
  12. Wu, Y., Li, R., Sun, S., Weile, J. & Roth, F. P. Improved pathogenicity prediction for rare human missense variants. Am. J. Hum. Genet. 108, 1891–1906 (2021). · DOI
  13. Amendola, L. M. et al. Performance of ACMG-AMP variant-interpretation guidelines among nine laboratories in the clinical sequencing exploratory research consortium. Am. J. Hum. Genet. 98, 1067–1076 (2016). · DOI
  14. Jordan, D. M. et al. Development and validation of a computational method for assessment of missense variants in hypertrophic cardiomyopathy. Am. J. Hum. Genet. 88, 183–192 (2011). · DOI
  15. Riesselman, A. J., Ingraham, J. B. & Marks, D. S. Deep generative models of genetic variation capture the effects of mutations. Nat. Methods 15, 816–822 (2018). · DOI · Paper page
  16. Toth-Petroczy, A. et al. Structured states of disordered proteins from genomic sequences HHS Public Access ETOC blurb. Cell 167, 158–170 (2016). · DOI
  17. Livesey, B. J. & Marsh, J. A. Using deep mutational scanning to benchmark variant effect predictors and identify disease mutations. Mol. Syst. Biol. 16, 1–12 (2020). · DOI
  18. Grimm, D. G. et al. The evaluation of tools used to predict the impact of missense variants is hindered by two types of circularity. Hum. Mutat. 36, 513–523 (2015). · DOI · Paper page
  19. Amendola, L. M. et al. Actionable exomic incidental findings in 6503 participants: challenges of variant classification. Genome Res. 25, 305–315 (2015). · DOI
  20. Cubuk, C. et al. Clinical likelihood ratios and balanced accuracy for 44 in silico tools against multiple large-scale functional assays of cancer susceptibility genes. Genet. Med. 23, 2096–2104 (2021). · DOI
  21. Ng, P. C. & Henikoff, S. SIFT: Predicting amino acid changes that affect protein function. Nucleic Acids Res. 31, 3812–3814 (2003). · DOI
  22. Sunyaev, S. R. et al. PSIC: profile extraction from sequence alignments with position-specific counts of independent observations. Protein Eng. 12, 387–394 (1999). · DOI
  23. Jordan, D. M. Predicting The Effects Of Missense Variation On Protein Structure, Function, And Evolution. ProQuest Dissertations and Theses. Vol. 132 (Harvard University, Graduate School of Arts & Sciences, 2015).
  24. Breen, M. S., Kemena, C., Vlasov, P. K., Notredame, C. & Kondrashov, F. A. Epistasis as the primary factor in molecular evolution. Nature 490, 535–538 (2012). · DOI
  25. Poelwijk, F. J., Socolich, M. & Ranganathan, R. Learning the pattern of epistasis linking genotype and phenotype in a protein. Nat. Commun. 10, 4213 (2019). · DOI
  26. Kondrashov, A. S., Sunyaev, S. & Kondrashov, F. A. Dobzhansky-Muller incompatibilities in protein evolution. Proc. Natl Acad. Sci. USA 99, 14878–14883 (2002). · DOI
  27. Sundaram, L. et al. Predicting the clinical impact of human mutation with deep neural networks. Nat. Genet. 50, 1161–1170 (2018). · DOI
  28. Davis, J. E., Voisine, C. & Craig, E. A. Intragenic suppressors of Hsp70 mutants: interplay between the ATPase- and peptide-binding domains. Proc. Natl Acad. Sci. USA 96, 9269–9276 (1999). · DOI
  29. Izumi, T. et al. Intragenic suppression of an active site mutation in the human apurinic/apyrimidinic endonuclease. J. Mol. Biol. 287, 47–57 (1999). · DOI
  30. Chen, R., Grobler, J. A., Hurley, J. H. & Dean, A. M. Second-site suppression of regulatory phosphorylation in Escherichia coli isocitrate dehydrogenase. Protein Sci. 5, 287–295 (1996). · DOI
  31. Marín, Ò., Aguirre, J. & de la Cruz, X. Compensated pathogenic variants in coagulation factors VIII and IX present complex mapping between molecular impact and hemophilia severity. Sci. Rep. 9, 9538 (2019). · DOI
  32. Baresić, A., Hopcroft, L. E. M., Rogers, H. H., Hurst, J. M. & Martin, A. C. R. Compensated pathogenic deviations: analysis of structural effects. J. Mol. Biol. 396, 19–30 (2010). · DOI
  33. Klein, G. & Georgopoulos, C. Identification of important amino acid residues that modulate binding of Escherichia coli GroEL to its various cochaperones. Genetics 158, 507–517 (2001). · DOI
  34. Tóth-Petróczy, A. & Tawfik, D. S. Slow protein evolutionary rates are dictated by surface-core association. Proc. Natl Acad. Sci. USA 108, 11151–11156 (2011). · DOI
  35. Sunyaev, S., Ramensky, V. & Bork, P. Towards a structural basis of human non-synonymous single nucleotide polymorphisms. Trends Genet. 16, 198–200 (2000). · DOI
  36. Stenson, P. D. et al. Human Gene Mutation Database (HGMD): 2003 update. Hum. Mutat. 21, 577–581 (2003). · DOI
  37. Karczewski, K. J. et al. The mutational constraint spectrum quantified from variation in 141,456 humans. Nature 581, 434–443 (2020). · DOI · Paper page
  38. Su Jung, K. et al. KRGDB: the large-scale variant database of 1722 Koreans based on whole genome sequencing Citation details. Database 2019, 146 (2018).
  39. Tadaka, S. et al. 3.5KJPNv2: an allele frequency panel of 3552 Japanese individuals including the X chromosome. Hum. Genome Var. 18, 6–28 (2019).
  40. Marks, D. S., Hopf, T. A. & Sander, C. Protein structure prediction from sequence variation. Nat. Biotechnol. 30, 1072–1080 (2012). · DOI · Paper page
  41. Fraley, C. & Raftery, A. E. Model-based clustering, discriminant analysis, and density estimation. J. Am. Stat. Assoc. 97, 611–631 (2002). · DOI
  42. Mistry, J. et al. Pfam: the protein families database in 2021. Nucleic Acids Res. 49, D412–D419 (2021). · DOI · Paper page
  43. Graham, W. J. 5th, Putnam, C. D. & Kolodner, R. D. The properties of Msh2-Msh6 ATP binding mutants suggest a signal amplification mechanism in DNA mismatch repair. J. Biol. Chem. 293, 18055–18070 (2018). · DOI
  44. Friedman, J. H. Greedy function approximation: a gradient boosting machine. Ann. Stat. 29, 1189–1232 (2001). · DOI · Paper page
  45. Friedman, J., Tibshirani, R. & Hastie, T. Additive logistic regression: a statistical view of boosting (With discussion and a rejoinder by the authors). Ann. Stat. 28, 337–407 (2000). · DOI · Paper page
  46. Livesey, B. J. & Marsh, J. A. Interpreting protein variant effects with computational predictors and deep mutational scanning. Dis. Model. Mech. 15, dmm049510 (2022). · DOI
  47. Reeb, J., Wirth, T. & Rost, B. Variant effect predictions capture some aspects of deep mutational scanning experiments. BMC Bioinform. 21, 107 (2020). · DOI
  48. Findlay, G. M. et al. Accurate classification of BRCA1 variants with saturation genome editing. Nature 562, 217–222 (2018). · DOI · Paper page
  49. Fayer, S. et al. Closing the gap: systematic integration of multiplexed functional data resolves variants of uncertain significance in BRCA1, TP53, and PTEN. Am. J. Hum. Genet. 108, 2248–2258 (2021). · DOI
  50. Livesey, B. J. & Marsh, J. A. Updated benchmarking of variant effect predictors using deep mutational scanning. bioRxiv https://doi.org/10.1101/2022.11.19.517196 (2022). · DOI
  51. Kotler, E., Shani, O., Marks, D. S., Oren, M. & Segal, E. A systematic p53 mutation library links differential functional impact to cancer mutation pattern and evolutionary conservation. Mol. Cell 71, 178–190 (2018). · DOI
  52. Matreyek, K. A. et al. Multiplex assessment of protein variant abundance by massively parallel sequencing. Nat. Genet. 50, 874–882 (2018). · DOI · Paper page
  53. Jia, X. et al. Massively parallel functional testing of MSH2 missense variants conferring Lynch syndrome risk. Am. J. Hum. Genet. 108, 163–175 (2021). · DOI
  54. Vaser, R., Adusumalli, S., Leng, S. N., Sikic, M. & Ng, P. C. SIFT missense predictions for genomes. Nat. Protoc. 11, 1–9 (2016). · DOI · Paper page
  55. Raimondi, D. et al. DEOGEN2: prediction and interactive visualization of single amino acid variant deleteriousness in human proteins. Nucleic Acids Res. 45, 201–206 (2017). · DOI
  56. Alver, M. et al. Recall by genotype and cascade screening for familial hypercholesterolemia in a population-based biobank from Estonia. Genet. Med. 21, 1173–1180 (2019). · DOI
  57. Alirezaie, N., Kernohan, K. D., Hartley, T., Majewski, J. & Hocking, T. D. ClinPred: prediction tool to identify disease-relevant nonsynonymous single-nucleotide variants. Am. J. Hum. Genet. 103, 474–483 (2018). · DOI
  58. Dines, J. N. et al. Systematic misclassification of missense variants in BRCA1 and BRCA2 “coldspots.”. Genet. Med. 22, 825–830 (2020). · DOI
  59. Sy, S. M. H., Huen, M. S. Y. & Chen, J. PALB2 is an integral component of the BRCA complex required for homologous recombination repair. Proc. Natl Acad. Sci. USA 106, 7155–7160 (2009). · DOI
  60. Chicco, D. & Jurman, G. The advantages of the Matthews correlation coefficient (MCC) over F1 score and accuracy in binary classification evaluation. BMC Genomics 21, 1–13 (2020). · DOI · Paper page
  61. Phan, L. et al. ALFA: Allele Frequency Aggregator. (National Center for Biotechnology Information, U.S. National Library of Medicine, 2020) www.ncbi.nlm.nih.gov/snp/docs/gsr/alfa/.
  62. Liu, X., Li, C., Mou, C., Dong, Y. & Tu, Y. dbNSFP v4: a comprehensive database of transcript-specific functional predictions and annotations for human nonsynonymous and splice-site SNVs. Genome Med. 12, 103 (2020). · DOI
  63. Erdős, G. & Dosztányi, Z. Analyzing protein disorder with IUPred2A. Curr. Protoc. Bioinform. 70, e99 (2020). · DOI
  64. Hopf, T. A. et al. The EVcouplings Python framework for coevolutionary sequence analysis. Bioinformatics 35, 1582–1584 (2019). · DOI
  65. Scrucca, L., Fop, M., Murphy, T. B. & Raftery, A. E. mclust 5: clustering, classification and density estimation using gaussian finite mixture models. R. J. 8, 289–317 (2016). · DOI
  66. Schwarz, G. Estimating the dimension of a model. Ann. Stat. 6, 461–464 (1978). · DOI · Paper page
  67. Mészáros, B., Erdos, G. & Dosztányi, Z. IUPred2A: context-dependent prediction of protein disorder as a function of redox state and protein binding. Nucleic Acids Res. 46, W329–W337 (2018). · DOI

Metadata

Authors
Luppino, F., Adzhubei, I. A., Cassa, C. A., & Toth-Petroczy, A.
Affiliations
  • Center for Systems Biology Dresden (DE)
  • Brigham and Women's Hospital (US)
Corresponding author
Cassa, Christopher A.
ORCID
Venue
Nature Communications
Publisher
Springer Science and Business Media LLC
Published
2023-04-19
Volume
14
Issue
1
Pages
2230-2230
ISSN
2041-1723 (ISSN-L)
DOI
10.1038/s41467-023-37661-z
PMC
PMC10115847
OpenAlex
W4366405805
Cited by
12
References
5
Citations by year
2026: 3 · 2025: 5 · 2024: 4
FWCI
1.99
Type
journal article
Language
en
Open access
Licence
Locations
  • Nature Communications (publishedVersion, cc-by)
  • bioRxiv (Cold Spring Harbor Laboratory) (acceptedVersion)
  • PubMed Central (publishedVersion, cc-by)
  • PubMed (publishedVersion)
  • RePEc: Research Papers in Economics (submittedVersion)
  • DOAJ (DOAJ: Directory of Open Access Journals) (submittedVersion, cc-by-sa)
Funders
  • Max-Planck-Gesellschaft (MPRGL)
  • U.S. Department of Health & Human Services | NIH | National Human Genome Research Institute (R01-HG010372)
Topics
  • Life Sciences › Biochemistry, Genetics and Molecular Biology › Genetics › Genomics and Rare Diseases
  • Life Sciences › Biochemistry, Genetics and Molecular Biology › Genetics › Genetic Associations and Epidemiology
  • Life Sciences › Biochemistry, Genetics and Molecular Biology › Cancer Research › Cancer Genomics and Diagnostics
Keywords
variant effect prediction, missense mutations, variants of uncertain significance, epistasis, protein structure, evolutionary conservation, variant interpretation
MeSH terms
Genetic Testing, Humans, United States, Genetic Variation, Mutation, Missense, Genomics

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review