Extension of Quality Metrics Display within VarFish for IVD Technical Validation

[your-highness]

In relation to varfish-org/varfish-server#528 I would like to discuss on the inclusion of quality metrics for a defined list of genes, _e.g. indication-based list of genes (In-Vitro Diagnostic). This facilitates the requirements of a "technical sample validation" within VarFish for usage in an IVD-regulated laboratory setting.

---

In a IVD-regulated diagnostic setting there exist several levels of quality and performance control even before variant data is cleared for technical & medical evaluation, e.g. DNA integrity, fragment length, # reads, ... (see risk assessment). While these control checks are crucial and mostly not needed after approval, a documentation of some quality metrics "next" to the results is needed for the technical & medical validation. For an IVD crucial quality metrics are:

• number of reads
• mean / median coverage
• mean / median fragment length (PE data)
• number of variants within IVD regions
• % of bases with 1x, 10x, 20x & 30x
• Enlistment of design regions lacking coverage, e.g. drop out regions (crucial to cover a fully classified IVD technical validation)

---

Currently, VarFish ingests quality metrics on the level of genome-wide alignment and variant statistics. Also downloads of TSV-formatted Variant Quality Control provided. These metrics are very helpful to judge an experiment successful.

In addition to these genome-wide statistics, VarFish could provide a way to ingest sub-panel specific quality criteria for display and download.

Unfortunately (and most importantly), VarFish does not provide a way to identify low-coverage design regions, e.g. CDS regions in an exome. For WES the list of low-coverage design regions might be long but a low-coverage warning for OMIM or ACMG defined lists could be raised.

For IVDs a listing of low/no-coverage design regions is crucial for technical validation and to estimate diagnostic sensitivity. Because IVDs only contain a few dozen to hundred genes a fully list of low-coverage regions could be possible.

---

P.S.: Invitation to @xiamaz to add or complement 😄