New in emedgene 30 (January 8th 2023)
New AI Models
This version was primarily focused on introducing new AI innovation, with the goal of streamlining workflows and further reducing time-per-case.
Shortlist CNVs and Compound Heterozygous SNV-CNV Variants
Our AI Shortlist now compiles CNVs with evidence, along with compound heterozygous SNV-CNV variants. We expect this exciting new capability to drive significant time savings.
The model expands our core AI Shortlist model to CNV variants, and similarly considers dozens of features used by geneticists to interpret a case. The model was validated in a cohort of 51 singletons previously solved by a CNV variant. In 92% of cases (n=47 / 51) the solving variant was identified in a short list of 10 variants. In 6% (n=3), the solving variant was present in the candidate list. Only 2% (n=1), or a single case, was missed by the new AI Shortlist.
Due to the challenging quality nature of CNV variants, the conventional successful AI Shortlist model for SNV is not affected by newly added CNV results. CNV and compound heterozygous variants shortlisted variants are cumulative to SNV/indels, and the default short list for your organization will be extended by 5 variants on average.
The model supports deletions and duplications called both by the DRAGEN CNV read-depth caller and DRAGEN SV caller.
Improved Phenotypic Match Model
We have a new and improved phenotypic match model, Phenomeld. Phenotypic match algorithms estimate the concordance between patient phenotypes and an associated disease, and are a critical component in the AI Shortlist model.
Phenomeld, which achieves an AUC of 0.94, is used both as part of the AI Shortlist model, and as an independent filter in the analysis tools. The improvement was achieved with ensemble machine learning, which utilizes several phenotypic match models to improve model performance.
The Phenomeld model was validated against two data sets.
Utilizing an internal dataset of 314 well characterized cases, Phenomeld was compared to the open source phen2gene model, as well as to the previous Emedgene Phenomatch model, outperforming both.
Internal Validation Set
A second validation used a set of 4648 ClinVar cases, and compared Phenomeld with phen2gene, where Phenomeld demonstrated improved results.
ClinVar Validation Set
For exploration of phenotypic match results outside of the AI Shortlist models, you can use the Phenomeld filter in the analysis tools.
Support for DRAGEN 4.0
This version supports DRAGEN 4.0 with ML and graph mapper, both for customers using the Emedgene DRAGEN pipeline and customers bringing their own DRAGEN.
DRAGEN 4.0 improves accuracy and reduces analysis time by ~26%.
Learn more about the DRAGEN 4.0 release.
Workflow Improvements
Related Cases Enrichment
The Related Cases component in the variant page has been enriched with interpretation and classification data, in order to streamline use of lab curated data.
The table now includes: solved/unsolved status, zygosity and affected and overlap % (for CNVs only).
When a case is selected, additional data is displayed: interpretation, phenotypes, proband ID, demographics, ACMG tags and case type.
Bulk Translation of Reports
Our reporting system now enables customers to upload custom data sets to reports. This enables customizable and flexible bulk translation and augmentation of reports with proprietary data.
Emedgene Curate Improvements
Private Networks now support CNV variants, with selectable % overlap.
Store gene interpretations and auto-populate to reports
New gene metrics card for easier gene curation
Enhancements
Lab Page | Gene Coverage - Region coverage statistics will now be calculated based on the case’s kit associated BED file. If none is defined, a default clinical exome BED will be used.
Variant Page | ACMG - You can now write and store notes on ACMG tags for SNV variants (previously available only for CNV variants).
Analysis Tools | Presets - New search bar in gene lists, for faster addition of gene lists to preset filters.
In organization settings, you can now view and manage Enrichment Kits.
Standard users can view the BED file used for any kit
Users with permissions can upload new BED files for any reference, both for existing and new kits. Please follow the file format guide for BED files (link accessible only to Emedgene platform users).
PolyPhen2 is back! Illumina has acquired a license to include this data in tertiary analysis platforms.
Gene Lists | When creating a gene list, the platform will save the NCBI ID for any gene in order to maintain compatibility with future gene symbol updates without the need for a lab gene list audit.
Expanded ethnicity list to support an increasingly global and diverse customer base.
Last updated
