CoChat
Sign Up
PreprintCited by 1Genomics and Rare DiseasesProtein Structure and DynamicsMachine Learning in Bioinformatics

The DOI matches its Crossref record; 1 of the 1 references it cites were found in a registry or an index.

ASCARIS: Positional Feature Annotation and Protein Structure-Based Representation of Single Amino Acid Variations

Open access
Cankara, Fatma, Doğan, Tunca·2022·Computational and Structural Biotechnology Journal
DOI 10.1101/2022.11.03.514934PMC PMC10562615

Preprint — published as 10.1016/j.csbj.2023.09.017 · Paper page

Open this paper in CoChat — free accountOpen PDF

Abstract

Motivation: Genomic variations may cause deleterious effects on protein functionality and perturb biological processes. Elucidating the effects of variations is critical for developing novel treatment strategies for diseases of genetic origin. Computational approaches have been aiding the work in this field by modeling and analyzing the mutational landscape. However, new approaches are required, especially for accurate and comprehensive representation and data-centric analysis of sequence variations. Results: In this study, we propose ASCARIS (Annotation and StruCture-bAsed RepresentatIon of Single amino acid variations - SAVs), a method for the featurization (i.e., quantitative representation) of SAVs, which could be used for a variety of purposes, such as predicting their functional effects or building multi-omics-based integrative models. In ASCARIS representations, we incorporated the correspondence between the location of the SAV on the sequence and 30 different types of positional feature annotations (e.g., active/lipidation/glycosylation sites; calcium/metal/DNA binding, inter/transmembrane regions, etc.) from UniProt, along with structural features such as protein domains, the location of variation (e.g., core/interface/surface), and the change in physico-chemical properties using models from PDB and AlphaFold-DB. We also mapped the mutated and annotated residues to the 3-D plane and calculated the spatial distances between them in order to account for the functional changes caused by variations in positions close to the functionally essential ones. Finally, we constructed a 74-dimensional feature set to represent each SAV in a dataset composed of ∼100,000 data points. We statistically analyzed the relationship between each of these features and the consequences of variations, and found that each of them carries information in this regard. To investigate potential applications of ASCARIS, we trained variant effect predictor models that utilize our SAV representations as input. We carried out both an ablation study and a comparison against the state-of-the-art methods over well-known benchmark datasets. We observed that our method displays a competing performance against widely-used predictors. Also, our predictions were complementary to these methods which is probably due to fact that ASCARIS has a rather unique focus in modeling variations. ASCARIS can be used either alone or in combination with other approaches, to universally represent SAVs from a functional perspective. Availability and implementation: The source code, datasets, results, and user instructions of ASCARIS are available at https://github.com/HUBioDataLab/ASCARIS .

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 · 10.1101/2022.11.03.514934

Cites

1 of 1 references verified0 citation errors · 0 not checkable

Statistics

Figures

Graphical abstract
Graphical abstract CC BY-NC-ND · PubMed Central
Fig. 1 The overall workflow of the proposed variant featurization/representation method, ASCARIS, together with its application to the problem of variant effect prediction.
Fig. 1 The overall workflow of the proposed variant featurization/representation method, ASCARIS, together with its application to the problem of variant effect prediction.CC BY-NC-ND · PubMed Central
Fig. 2 Dataset construction and featurization steps: (a) the sources of structural information and statistics about the data from each source, (b) the representation of different types of features on SAV representations, (c) mapping of positional sequence annotations and the SAV onto the 3-D protein structure, and part of the SAV feature vector that corresponds to these annotations.
Fig. 2 Dataset construction and featurization steps: (a) the sources of structural information and statistics about the data from each source, (b) the representation of different types of features on SAV representations, (c) mapping of positional sequence annotations and the SAV onto the 3-D protein structure, and part of the SAV feature vector that corresponds to these annotations.CC BY-NC-ND · PubMed Central
Fig. 3 . (a) Distribution of neutral and deleterious SAV data points according to their domain region correspondence. Mutations found within the domains tend to be more deleterious (61.3%) compared to the ones outside (30.6%), (b) distribution of neutral and deleterious SAV data points according to their location on the structure of the protein. Mutations found within the core and interface regions tend to be more deleterious (71.7% and 69.4%, respectively) compared to the ones on the surface (48.6%).
Fig. 3 . (a) Distribution of neutral and deleterious SAV data points according to their domain region correspondence. Mutations found within the domains tend to be more deleterious (61.3%) compared to the ones outside (30.6%), (b) distribution of neutral and deleterious SAV data points according to their location on the structure of the protein. Mutations found within the core and interface regions tend to be more deleterious (71.7% and 69.4%, respectively) compared to the ones on the surface (48.6%).CC BY-NC-ND · PubMed Central
Fig. 4 . (a) The coverage of each annotation category on the proteins in our SAV dataset (e.g., nearly 23% of proteins in our dataset have at least one active site annotation in UniProtKB), (b) the rates of deleterious vs neutral variations for each annotation category, which are calculated considering the SAV data points in our dataset that coincide with a positional annotation on the same residue, (c) Mean spatial distances between annotated residues and mutated residues in the structure, calculated independently for each annotation category, considering the cases in which the SAV and annotated residue do not coincide on the same residue. In the case of the existence of multiple annotations on a protein, only the annotation that is located spatially closest to the mutation of interest is taken into account. Dashed lines indicate overall averages calculated by taking the mean of all distances. For the calculation of Euclidean distances between residues, the coordinates of C⍺ of the respective amino acids are extracted from PDB models.
Fig. 4 . (a) The coverage of each annotation category on the proteins in our SAV dataset (e.g., nearly 23% of proteins in our dataset have at least one active site annotation in UniProtKB), (b) the rates of deleterious vs neutral variations for each annotation category, which are calculated considering the SAV data points in our dataset that coincide with a positional annotation on the same residue, (c) Mean spatial distances between annotated residues and mutated residues in the structure, calculated independently for each annotation category, considering the cases in which the SAV and annotated residue do not coincide on the same residue. In the case of the existence of multiple annotations on a protein, only the annotation that is located spatially closest to the mutation of interest is taken into account. Dashed lines indicate overall averages calculated by taking the mean of all distances. For the calculation of Euclidean distances between residues, the coordinates of C⍺ of the respective amino acids are extracted from PDB models.CC BY-NC-ND · PubMed Central
Fig. 5 Performance results (AUROC) of ASCARIS and 46 different VEPs from the literature, on the deep mutational scanning data. AUROC scores indicate the success of methods in distinguishing deleterious (pathogenic) SAVs from the neutral (benign) ones on the; (a) BRCA1, (b) CALM1, and (c) P53 genes. The number of test SAVs for each class are shown in plot titles. ASCARIS scores are shown in blue colored bars (DMS: deep mutational scanning, Supervised: machine learning -ML- models that learns from labeled samples, Unsupervised: ML models that learns from evolutionary conservation and multiple sequence alignments, Empirical: different from ML models, these are VEPs that rely on empirical calculations, Metapredictor: mostly supervised ML models that utilize the results of other VEPs as its input).
Fig. 5 Performance results (AUROC) of ASCARIS and 46 different VEPs from the literature, on the deep mutational scanning data. AUROC scores indicate the success of methods in distinguishing deleterious (pathogenic) SAVs from the neutral (benign) ones on the; (a) BRCA1, (b) CALM1, and (c) P53 genes. The number of test SAVs for each class are shown in plot titles. ASCARIS scores are shown in blue colored bars (DMS: deep mutational scanning, Supervised: machine learning -ML- models that learns from labeled samples, Unsupervised: ML models that learns from evolutionary conservation and multiple sequence alignments, Empirical: different from ML models, these are VEPs that rely on empirical calculations, Metapredictor: mostly supervised ML models that utilize the results of other VEPs as its input).CC BY-NC-ND · 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: 95

  1. 1971, An empirical hydrophobicity scale for α-amino-acids and some of its applications, International Journal of Biochemistry, 2, 537, 10.1016/0020-711X(71)90023-1 · DOI
  2. 2013, Predicting functional effect of human missense mutations using PolyPhen-2, Curr. Protoc. Hum. Genet, 76, 7
  3. 10.1038/nmeth0410-248 · DOI · Paper page
  4. 2014, PredictSNP: Robust and Accurate Consensus Classifier for Prediction of Disease-Related Mutations, PLoS Computational Biology, 10, e1003440, 10.1371/journal.pcbi.1003440 · DOI
  5. 10.1093/nar/28.1.235 · DOI · Paper page
  6. 10.1007/s10585-009-9253-z · DOI
  7. 2001, Machine Learning, 45, 261, 10.1023/A:1017934522171 · DOI
  8. 10.1093/nar/gkm238 · DOI
  9. 10.1093/bioinformatics/btn435 · DOI
  10. 10.1002/humu.21047 · DOI
  11. 10.1186/1471-2164-14-1 · DOI · Paper page
  12. 10.1093/nar/gki375 · DOI
  13. 2019, Integrating molecular networks with genetic variant interpretation for precision medicine, Wiley Interdiscip. Rev. Syst. Biol. Med, 11, e1443, 10.1002/wsbm.1443 · DOI
  14. 10.1158/0008-5472.CAN-09-1133 · DOI
  15. 10.1371/journal.pone.0236962 · DOI
  16. 10.1093/bioinformatics/btv195 · DOI
  17. 10.1093/bioinformatics/bth029 · DOI
  18. 10.1038/nature15393 · DOI · Paper page
  19. 10.1093/nar/gky1049 · DOI · Paper page
  20. Cox, M.A.A. and Cox, T.F. Multidimensional Scaling. Handbook of Data Visualization 2008:315–347. · DOI
  21. Cristianini, N. and Shawe-Taylor, J. An Introduction to Support Vector Machines and Other Kernel-based Learning Methods. 2000. · DOI · Paper page
  22. 10.1002/gepi.20642 · DOI
  23. 2015, Functional and Structural Consequences of Damaging Single Nucleotide Polymorphisms in Human Prostate Cancer Predisposition Gene RNASEL, Biomed Res Int, 2015, 271458
  24. 2019, 3D spatial organization and network-guided comparison of mutation profiles in Glioblastoma reveals similarities across patients, PLoS computational biology, 15, e1006789, 10.1371/journal.pcbi.1006789 · DOI
  25. 10.7717/peerj.5298 · DOI
  26. 10.1093/bioinformatics/btw114 · DOI
  27. 2021, Protein domain-based prediction of drug/compound–target interactions and experimental validation on LIM kinases, PLoS Computational Biology, 17, e1009171, 10.1371/journal.pcbi.1009171 · DOI
  28. 2021, CROssBAR: comprehensive resource of biomedical relations with knowledge graph representations, Nucleic Acids Research, 49, e96, 10.1093/nar/gkab543 · DOI
  29. Engin, H.B. , Hofree, M. and Carter, H. Identifying mutation specific cancer pathways using a structurally resolved protein interaction network. Pac. Symp. Biocomput. 2015:84–95. · DOI
  30. 10.1038/nature13835 · DOI · Paper page
  31. 10.1093/bioinformatics/btv291 · DOI
  32. 10.1006/jcss.1997.1504 · DOI · Paper page
  33. Friedman, J. , Hastie, T. and Tibshirani, R. Additive logistic regression: a statistical view of boosting (With discussion and a rejoinder by the authors). The Annals of Statistics 2000;28(2). · DOI · Paper page
  34. 1995, Learning binary relations using weighted majority voting, Machine Learning, 20, 245, 10.1007/BF00994017 · DOI
  35. 10.1016/0022-5193(73)90075-1 · DOI
  36. 10.1016/j.ajhg.2011.03.004 · DOI
  37. 10.1126/science.185.4154.862 · DOI
  38. 10.1002/humu.22768 · DOI · Paper page
  39. 10.1073/pnas.0403255101 · DOI
  40. 10.1038/10297 · DOI
  41. Hastie, T. , Friedman, J. and Tibshirani, R. The Elements of Statistical Learning. Springer Series in Statistics 2001. · DOI · Paper page
  42. 10.1186/s12859-015-0611-3 · DOI
  43. 10.1073/pnas.0903103106 · DOI · Paper page
  44. 10.1016/j.molp.2016.06.004 · DOI
  45. 10.1093/nar/gkm405 · DOI
  46. 10.1002/humu.20534 · DOI
  47. 10.1093/nar/27.1.355 · DOI
  48. 10.1038/nrg.2015.17 · DOI
  49. 10.1093/bioinformatics/btz595 · DOI
  50. 10.1093/nar/gkx1153 · DOI · Paper page
  51. 10.1172/JCI34772 · DOI · Paper page
  52. 10.1093/hmg/ddw091 · DOI
  53. 10.1002/humu.23738 · DOI
  54. 10.1038/nmeth.4540 · DOI
  55. 10.1093/nar/gky1100) · DOI · Paper page
  56. 2016, FreeSASA: An open source C library for solvent accessible surface area calculations, F1000Res, 5, 189, 10.12688/f1000research.7931.1 · DOI
  57. 10.1186/1471-2105-9-357 · DOI
  58. 10.1093/bioinformatics/btq028 · DOI
  59. 10.1093/bioinformatics/btn654 · DOI
  60. 10.1093/nar/gkg509 · DOI
  61. 10.1371/journal.pone.0066273 · DOI
  62. 2021, Computational evaluation of interactions between olfactory receptor OR2W1 and its ligands, Genomics Inform, 19, e9, 10.5808/gi.21026 · DOI
  63. 10.1002/pro.3774 · DOI
  64. 10.1093/nar/gkt1144 · DOI
  65. 10.1093/bioinformatics/btt691 · DOI · Paper page
  66. 10.1016/j.cell.2015.02.029 · DOI · Paper page
  67. 10.1093/bioinformatics/btw361 · DOI
  68. 10.1093/nar/gky1016 · DOI · Paper page
  69. 10.1186/gb-2007-8-11-r232 · DOI
  70. 10.1093/nar/gkr407 · DOI
  71. 10.1002/prot.25416 · DOI
  72. 10.1002/humu.22204 · DOI
  73. 10.1038/nrg3051 · DOI · Paper page
  74. 10.1093/bioinformatics/btw192 · DOI · Paper page
  75. 10.1038/nmeth.2890 · DOI · Paper page
  76. 10.1093/nar/gki387 · DOI · Paper page
  77. 10.1002/ppul.22567 · DOI
  78. 10.1002/humu.22225 · DOI
  79. 10.1101/gr.3804205 · DOI
  80. 10.1016/j.cell.2014.01.051 · DOI
  81. 10.1002/humu.22093 · DOI
  82. 10.1002/humu.20896 · DOI
  83. 10.1136/jmg.2005.033878 · DOI
  84. 10.1002/humu.21445 · DOI · Paper page
  85. 10.1093/protein/10.1.7 · DOI
  86. 10.1002/bies.201600062 · DOI
  87. 10.1038/s42256-022-00457-9 · DOI
  88. 10.1007/s10479-011-0841-3 · DOI
  89. 10.1093/nar/gky427 · DOI
  90. 10.1093/nar/gkr363 · DOI
  91. 1990, In vitro gamma-carboxylation of a 59-residue recombinant peptide including the propeptide and the gamma-carboxyglutamic acid domain of coagulation factor IX. Effect of mutations near the propeptide cleavage site, J. Biol. Chem, 265, 13124, 10.1016/S0021-9258(19)38275-4 · DOI
  92. 10.1007/s00726-012-1407-7 · DOI
  93. 10.1016/j.jmb.2005.08.020 · DOI
  94. 10.1016/j.jmb.2005.12.025 · DOI
  95. 2018, Unraveling the causes of adaptive benefits of synonymous mutations in TEM-1 beta-lactamase, Heredity (Edinb), 121, 406, 10.1038/s41437-018-0104-z · DOI

Metadata

Authors
Cankara, F., & Doğan, T.
Affiliations
  • Koç University (TR)
  • Hacettepe University (TR)
Corresponding author
Doğan, Tunca
ORCID
Venue
Computational and Structural Biotechnology Journal
Publisher
openRxiv
Published
2022-11-04
Volume
21
Pages
4743-4758
ISSN
2001-0370 (ISSN-L)
DOI
10.1101/2022.11.03.514934
PMC
PMC10562615
OpenAlex
W4386813159
Cited by
1
References
1
Citations by year
2026: 1
FWCI
0.17
Type
journal article
Language
en
Open access
Licence
cc-by-nc-nd
Locations
  • Computational and Structural Biotechnology Journal (publishedVersion, cc-by-nc-nd)
  • bioRxiv (Cold Spring Harbor Laboratory) (acceptedVersion, cc-by)
  • PubMed Central (publishedVersion, cc-by-nc-nd)
  • PubMed (publishedVersion)
  • Digital Collections portal (Koç University) (submittedVersion)
  • DOAJ (DOAJ: Directory of Open Access Journals) (submittedVersion, cc-by-sa)
Funders
Koç Üniversitesi
Topics
  • Life Sciences › Biochemistry, Genetics and Molecular Biology › Genetics › Genomics and Rare Diseases
  • Life Sciences › Biochemistry, Genetics and Molecular Biology › Molecular Biology › Protein Structure and Dynamics
  • Life Sciences › Biochemistry, Genetics and Molecular Biology › Molecular Biology › Machine Learning in Bioinformatics
Keywords
variant effect prediction, protein structure, functional annotation, protein binding sites, physicochemical properties, multi-omics integration, Ascaris

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