Computational evidence plays a vital role in the interpretation of variants using the ACMG guidelines. This includes functional prediction scores like SIFT and PolyPhen2, as well as conservation metrics such as GERP++ and PhyloP. In this webcast, we review the conservation scores and functional prediction algorithms available in VSClinical. This includes a discussion of our own implementation of these algorithms, along with a comparison to more recent variant prediction methods.

1. Functional Predictions and Conservation
Scores in VSClinical
20 most promising
Biotech Technology
Providers
Top 10 Analytics
Solution Providers
Hype Cycle for
Life sciences

2. Q & A
Please enter your questions into your GoToWebinar Panel

3. Thanks to NIH & Stakeholders
▪ NIH Grant Supported
- Research reported in this publication
was supported by the National Institute
Of General Medical Sciences of the
National Institutes of Health under
Award Number R43GM128485. PI is Dr.
Andreas Scherer, CEO Golden Helix.
The content is solely the responsibility
of the authors and does not necessarily
represent the official views of the
National Institutes of Health.
▪ ACMG Guidelines Author Collaborator:
- Dr. Elaine Spector (Childrens Colorado, USA)
▪ Stakeholders:
- Dr. Abdallah Elias (Shodair Children’s Hospital, USA)
- Dr. Ahmed Alfares, King Abdul Aziz Medical City, Saudi
Arabia),
- Dr. Bailey Glen (Medical University of South Carolina,
USA)
- Dr. Jim Weber (PreventionGenetics, USA)
- Dr. Qin Hae and Dr. Line Larsen (Amplexa, Denmark)
- Dr. Val Hyland (Pathology Queensland, Australia).

4. Golden Helix – Who We Are
Golden Helix is a global
bioinformatics company founded
in 1998.
GWAS
Genomic Prediction
Large-N Population Studies
RNA-Seq
Large-N CNV-Analysis
Variant Warehouse
Centralized Annotations
Hosted Reports
Sharing and Integration
Variant Calling
Filtering and Annotation
Clinical Reports
CNV Analysis
Pipeline: Run Workflows

5. Cited in over 1200 peer-reviewed
publications

6. Over 350 customers globally

7. Golden Helix – Who We Are
When you choose a Golden Helix solution, you get more than just
software
▪ REPUTATION
▪ TRUST
▪ EXPERIENCE
▪ INDUSTRY FOCUS
▪ THOUGHT
LEADERSHIP
▪ COMMUNITY
▪ TRAINING
▪ SUPPORT
▪ RESPONSIVENES
S
▪ INNOVATION and
SPEED
▪ CUSTOMIZATIONS

8. Genetic Testing Process
Sample Prep Sequencing Align & Call Annotate
& Filter
Variant
Interpretation
Report
Sentieon
& VS-CNV
VarSeq VSReportsVSClinical
Golden Helix Clinical Suite

9. VSClinical
▪ Complete Support for ACMG Guideline
Workflow:
- Implements a guided workflow for following the ACMG guideline
scoring and classifying
- Place criteria into conceptually related groups, paired with their
opposites, and formatted as answerable question.
▪ Aggregate and Automate:
- Questions have supporting evidence presented with rich and
interactive visuals
- Automatically computed recommendations for questions that have
explicit bioinformatic evidence, with supporting reasons for each
answer.
▪ Expert and Beginner Friendly:
- Start with descriptive summaries and recommendations for a
variant
- Deep dive into Population Catalogs, Gene Impact, Published Studies
and Clinical tabs
- Integrated documentation, readings on scoring criteria and citations

10. Computational Evidence
▪ The ACMG guidelines utilize various in silico
methods as supporting evidence:
- Splice Site Prediction
- Functional Prediction
- Conservation Scores
▪ VSClinical enables easy evaluation of this
evidence

11. Functional Prediction Algorithms
▪ SIFT
- Uses matrix to encode the probability of each amino
acid at each position of the protein
- Probabilities computed from protein sequence
alignment
▪ PolyPhen2
- Naïve Bayes multi-evidence approach
- Uses similar protein alignment-based probability score
called PSIC
- Incorporates nine other metrics

12. Implementation and Comparison
▪ VSClinical supports both SIFT and PolyPhen2
- Original algorithms use alignments queried via PSI-BLAST
from a database such as SWISS-PROT
- VSClinical utilizes UCSC’s 100-way alignment instead
▪ We compared our implementation to dbNSFP’s
precomputed scores
- Used ClinVar pathogenic and benign variants as ground truth
- Compared percentage of correctly classified pathogenic and
benign variants
▪ Compared against several more modern algorithms
- MutationTaster, MutationAccessor, FATHMM, and Provean

13. Comparison to dbNSFP precomputed scores
Accuracy SIFT (Vertebrates) PolyPhen2 (Vertebrates) SIFT (dbNSFP) PolyPhen2 (dbNSFP)
Pathogenic 94.40% 93.80% 90.33% 88.95%
Benign 59.10% 66.80% 66.65% 68.79%

14. Comparison to dbNSFP precomputed scores
Accuracy SIFT (Vertebrates)
PolyPhen2
(Vertebrates)
MutationTaster MutationAccessor FATHMM Provean
Pathogenic 94.40% 93.80% 84.72% 80.84% 95.41% 85.96%
Benign 59.10% 66.80% 58.59% 60.65% 74.00% 77.90%

15. Conservation Scores
▪ Conservation Scores
- Use phylogenetic models
- Find maximum likelihood scaling factor for
model given an alignment
▪ GERP++
- Uses rejected substitutions (RS) as test statistic
- RS value is computed from the neutral rate n and
the maximum likelihood scaling factor θ
RS = n(1- θ)
▪ PhyloP LRT
- Two times the difference in log likelihood
between
- null hypothesis (no scaling factor)
- alternative hypothesis (maximum likelihood
scaling factor)

16. PP3/BP4
▪ ACMG Criteria PP3/BP4
- Recommended when all computational evidence
supports a deleterious/tolerated effect
▪ VSClinical automatically recommend these
criteria based on
- Splice site prediction algorithms
- Functional predictions algorithms
- Conservation scores
▪ In ClinVar:
- PP3 is recommended for 99% of pathogenic variants
- BP4 is recommended for 70% of benign variants

17. [Demo in VarSeq]

18. Q & A
Please enter your questions into your GoToWebinar Panel