Rainbow is a proprietary CNV detection tool powered by advanced AI algorithms developed by Genoox, that enables CNV and Compound SNV-CNV analysis.
Rainbow by Franklin provides powerful CNV capabilities such as:
Copy Number Variants in small Panels / WES / WGS originating from Germline and Somatic samples
Detection down to two exons deletion or duplication (heterozygous) or one exon (homozygous) resolution
Rainbow creates a unique model by using a cohort of >30 samples from more than one batch of the same kit and laboratory - in order to reduce the batch effect
Once the model is defined, the workflow is fully automated from raw data to variant calling and interpretation
The model is frequently re-trained by using the latest samples for best optimization and accuracy
Franklin's in-house advanced AI algorithm "Rainbow"
Each exon’s predicted coverage is calculated by using over 50 unique "predictors"* whose coverage is statistically correlated to the exon’s coverage.
This process allows Rainbow to predict an exon’s or region’s copy number, and also to determine the prediction score, meaning how accurate the correlation is.
*predictors are different exons / regions, from various locations, preferably from different chromosomes.
The Confidence and CNV region viewer
Rainbow by Franklin classifies each CNV into Failed / Low / Medium / High confidence based on the Prediction score, Predicted copy number and number of following targets (exons or regions).
The variant confidence widget visualizes 3 AI-inferred confidence parameters:
Predicted copy number
The aggregate value for each parameter is listed on top. Clicking the score displays a detailed, per-exon view for:
The relevant sample - in blue
Family members’ samples (if exists) - in light blue
Additional reference samples - in grey
Additional reference samples can be added or filtered out from the display using the dropdown menu on the top-right.
Hovering over an exon point in the sample’s graph reveals the exact values for that exon.
The genomic range of the graph is shown on the top menu and allows navigation by changing the range itself, or using the navigation buttons (zoom-in/out, move left/right).
The median depth in the del/dup area averaged across the exons in that area.
An exon-level median depth of each sample can be observed by hovering the dots.
Prediction Score represents the potential calling accuracy of the CNV model in the variant region. This score is derived from the model itself, and therefore the same for all samples. An exon-level prediction score can be observed by hovering the dots.
A low prediction score is usually caused by one of the following reasons:
Low coverage correlation to predictor exons.
The entire gene is not included in the model due to low clinical significance.
Predicted copy number
The variant’s predicted copy number, as calculated by the model.
3 Heterozygous duplication
1 Heterozygous deletion
0 Homozygous deletion
An exon-level predicted copy number of each sample can be observed by hovering the dots.
Variant confidence assessment:
Deletion should consist of at least 2 consecutive exons for it to be considered as a reliable call.
The majority of exons included in the call should have:
A predicted copy number of at most 1.2, and a prediction score above 0.9, while the reference samples should have normal copy numbers (between 1.7 to 2.4).
When compared to the reference samples with similar coverage in exons adjacent to the variant region - the subject sample’s Median Coverage within the variant region should be at most 60% of the expected coverage
If the above holds and the variant’s predicted copy number is between 1.2 to 04, the deletion is considered heterozygous. If the variant’s predicted copy number is below 0.4 and the median coverage is negligible (<=6), deletion can be considered homozygous.
If the majority of exons have a predicted copy number above 1.4, prediction score below 0.8, or a significant number (>20%) of reference samples have abnormal copy numbers - the variant can be considered an artifact.
In any other case, variant call reliability should be considered uncertain.
Duplication should consist of at least 3 consecutive exons for it to be considered as a reliable call.
The majority of exons should have:
A predicted copy number of at least 2.8 and prediction score above 0.9, while reference samples should have normal copy numbers (between 1.7 to 2.4).
When compared to the reference samples with similar coverage in exons adjacent to the variant region - the subject sample’s Median Coverage within the variant region should be at least 140% of the expected coverage
If the majority of exons have a predicted copy number below 2.5, prediction score below 0.8, or a significant number (>20%) of reference samples have abnormal copy numbers - variant can be considered an artifact.
In any other case, variant call reliability should be considered uncertain
Rainbow utilizes Franklin's automated ACMG - CNV classification
Franklin's automated ACMG - CNV classification is based on the most recent ClinGen / ACMG guidelines (Sep. 2019)