Cervical cancer, caused by persistent infection with high risk (hr) HPV affects ~500,000 women globally and causes ~260,000 deaths annually. Although HPV immunisation has been successfully implemented it will take decades to see an effect of the new nonavalent vaccine. Highly sensitive hrHPV testing is likely to become the dominant primary screen. However, hrHPV infection is common and only a fraction of women are at risk of developing cancer. 40% of hrHPV+ women are cytology negative and triage by proposed adjunctive tests such as p16 (in conjunction with ki67) are insufficient. A molecular test is needed to identify clinically significant HPV infection.
We aim to identify and validate novel DNAm biomarkers which, in combination with our existing DNAm classifier1.2.3, will aid development of a new improved classifier for identification of high grade CIN and ultimately improve the current cervical cancer screening gold standard. To achieve this, we propose to measure DNAm of predefined sites in HPV16, 18, 31 and 33 in a set of 350 hrHPV+ women with normal cytology. In addition we will measure genome scale human DNAm with RRBS (Reduced representation bisulfide sequencing) followed by machine learning using MS-SPCA to identify ~100 sites which consistently appear in the best ranking models. Then, the best sites will be further sifted by additional multivariate data modelling to provide us with a minimum number of required classifier sites. Finally, these selected sites will be validated in a second set of 200 well characterised cervical samples.
DNA was purified from frozen LBC samples for methylation testing using Qiagen DNA extraction kits following standardized methods1.2.3. The RRBS method was designed to obtain 30 million reads per sample using 50bp SE reads corresponding to 20X depth coverage of each CpG. All specimens used in the project from the ARTISTIC and Scottish HPV Archive will be tested for HPV DNAm by pyrosequencing3. The HPV sites tested are selected according to established knowledge about their role and importance4.
The preliminary results indicate that this approach significantly improved risk classification tool to triage hrHPV+ women to colposcopy. The new classifier is expected to come close to the high sensitivity of current hrHPV tests (90-95%) but deliver a substantially higher specificity and PPV (both ~70%) than current molecular reflex tests (30-40% and 40-50% respectively).
This new algorithm would allow more efficient utilization of colposcopy services while hrHPV+ women negative for the triage classifier could be followed up at suitably frequent intervals to safely catch most, if not all, triage false negatives.
1. Vasiljevic N, et al. J Clin Virol 2014 ;59:161-6.
2. Brentnall AR, et al. Int J Cancer 2014 ;135:1425-32.
3. Vasiljevic N, et al. Gynecol Oncol 2014 ;132:709-14.
4. Sun C, et al. Gynecol Oncol 2011 ;121:59-63.