Table 1.

Considerations and key conclusions of this article

Considerations	Conclusions
Formalin-induced alterations to DNA	Modifications in FFPE-DNA mostly occur pseudorandomly, but more frequently in AT-rich genomic regions, leading to a higher prevalence of GC-rich sequences than in FF-DNA. Base modifications, inter-strand cross-linking, base excision, polydeoxyribose fragmentation and cytosine deamination, among others, constitute to the artefact repertoire of FFPE-DNA.
Consequences of formalin fixation	Formalin modifications are complex and still not completely understood. Artefacts can be mistaken as true variants, especially if their allelic frequency exceeds filter thresholds, which arises in regions of locally low coverage, which in turn is caused by reduced library complexity and non-uniform coverage.
Pre-analytical sample quality and its specifications (Parameter I)	Specimens of a decade or older can be considered if fixed in buffered formalin. Target tissue (e.g. tumour area) should be optimally enriched. FFPE-DNA extraction is critical and should be performed with caution. The average fragment length is an easy-to-determine metric that correlates with coverage uniformity, one of the most important quality criteria in FFPE-DNA sequencing.
Optional application of DNA repair treatment (Parameter II)	Repair should be considered especially for severely impaired specimens in smaller hypothesis-driven studies. Simple repair (e.g. UDG treatment) should be avoided in favour of BER-based enzymatic repair protocols. We recommend a BER-based repair protocol that restores fragments to increase coverage evenness and decrease artefact allele frequency.
Analytical sample preparation (Parameter III)	When FFPE-DNA is prepared for sequencing, individualised workflow adaptations can improve the outcome. Useful adaptations include higher DNA input amounts, mild shearing conditions or tagmentase, FFPE-specific kits, the use of UMIs, and replicate strategies.
Sequencing	In general, up to four-fold deeper sequencing is necessary than for undamaged FF-DNA. Very deep sequencing (e.g. 5000–7500×) is necessary if dual-UMI strand-specific error correction is the aim.
Bioinformatic analysis (Parameter IV)	Bioinformatic filters can facilitate the discrimination between true variants and artefacts. Different UMI filter and error correction strategies can drastically reduce artefacts but at the cost of coverage. This might impair the sensitivity in detecting true variants with low VAFs.
Assessing the rigor of published research	Not providing enough technical and methodological details limits the scientific quality and integrity of FFPE-studies. An adapted study design and providing a minimal set of information can improve the situation in the future. When developing mitigation strategies, the sole focus on artefact count reduction is too simplistic as the restoration of additional fragments might be more effective.