Objectives The S5 DNA-methylation classifier, based on target CpG sites of the human gene EPB41L3, and viral late gene regions of HPV-16,18,31 and 33 (Lorincz A et al., 2016) has demonstrated better performance for detection of CIN2/3-women than either HPV16/18 genotyping, cytology or combination. We tested the performance of S5 in detecting invasive cancers versus pre-cancers and quantified the degree of separation between normal/CIN1, CIN2/3 and invasive cancer S5 scores.

Methods Methylation status of the S5-CpGs was tested in DNA extracted from exfoliated cervical cell from the UK(n=138),Spain(n=100),Colombia(n=96),Philippines(n=50),Georgia(n=42) and Ethiopia(n=79). Samples were histologically defined as negative/CIN1, CIN2/3 and invasive cancer. DNA-bisulfite conversion was carried out and followed by pyrosequencing for the 6 components of S5. Average methylation was calculated for each marker to define the S5 score.

Results Methylation at all sites increased proportionally with disease severity showing a Cuzick-trend of z=9.2933(p<2.2x10–16). The separation of normal/CIN1 from CIN2/3 and from cancer was highly-significant (Mann-Whitney, all p<0.0001). S5 also showed highly-significant difference between CIN2/3 and invasive cancer from matched cohorts: UK(p<0.003), Spain(p<0.0001) and Colombia(p<0.003). ROC-curves were used to assess the diagnostic potential of S5 in differentiating cancers from CIN2/3. The AUC was 0.86(CI 95%:0.7965 to 0.9131,p<0.0001) with a sensitivity of 79.8% and a specificity of 83.1%, based on a cut-off at highest Youden J-index.

Conclusions The S5 methylation classifier may be useful in cervical screening programs for identifying progressive pre-cancers in women. Although the separation was very good, there is room for improvement by addition of new markers derived from our ongoing NGS multi-omics study.