Table 3 CNVoyant predictions for six diagnostic CNVs observed in the clinical setting. Displayed are CNVoyant predictions for six diagnostic CNVs identified in five patients. Each row corresponds to a CNV and provides the following information; Patient: the de-identified identifier for each patient; Diagnostic CNV: the chromosomal location (GRCh38) and type of CNV (deletion or duplication); CNV Pathogenic Rank: the prioritized rank assigned to the CNV among all of the patient’s CNVs, sorted by pathogenic probability; Prediction (Probability): CNVoyant probability predictions for the CNV being of benign, uncertain (VUS), or pathogenic clinical significance. The highest class prediction is indicated in bold on a per-variant basis.

From: CNVoyant a machine learning framework for accurate and explainable copy number variant classification

Patient

Diagnostic CNV

CNV pathogenic rank

Prediction (probability)

   

Benign

VUS

Pathogenic

1

chr22 : 20,327,301–21,112,300 (Deletion)

1

0.008

0.063

0.928

2

chr5 : 176,119,001–178,006,400 (Deletion)

1

0.010

0.011

0.979

3

chr10 : 48,125,369–50,707,368 (Deletion)

1

0.011

0.182

0.806

4

chr22 : 43,274,001–50,780,900 (Deletion)

1

0.002

0.005

0.994

 

chr4 : 86,599,601–88,823,600 (Deletion)

2

0.015

0.112

0.874

5

chrX : 150,194,201–155,699,000 (Duplication)

1

0.011

0.010

0.979