Abstract 1143P
Background
Genome-wide DNA methylation sequencing provides plenty of differentially methylated regions for further implementation with MSRE-qPCR. Multiplex design is challenging requiring unified PCR conditions for all primers/probes, high specificity for target regions and inclusion of MSRE recognition sites. To address these requirements, we developed a method to design multiplex panels of amplicons, and to estimate their diagnostic potential. We have applied this method to our genome-wide DNA methylation XmaI-RRBS data for triple-negative breast cancer (TNBC) to predict response to neoadjuvant chemotherapy (NAC).
Methods
XmaI-RRBS dataset for 34 TNBC biopsies taken prior to NAC was used. Virtual amplicons were designed under the following criteria: at least two MSRE (BstHHI and/or HpaII) recognition sites within the amplicon, the amplicon length no more than 100 bp to provide high MSRE-qPCR efficiency, the average difference in the methylation level between adjacent CpG pairs no more than 10%. To select amplicons, MSRE sites were hierarchically clusterized with the distance metric of physical distance over the genome and the difference in their methylation level, followed by complete-linkage agglomeration to prevent lengthening of amplicons via chaining phenomenon. Diagnostic potential was assessed with cross-validated AUC. Markers with AUC ≥ 75% were selected to form panels.
Results
Genes APCDD1L, RUSC1-AS1, MYO15B, EXOC2, THBS2, MXRA5, ANKRD64 were selected to form possible combinations of markers. Eventually, 120 combinations of amplicons panels that discriminate NAC response were obtained and top-10 are shown in the table. Table: 1143P
Top 10 combinations of amplicons that discriminate NAC responding and non-responding TNBC sorted in descending order by AUC. Three gene (RUSC1-AS1, MYO15B, ANKRD46) panel exhibits highest value for AUC, sensitivity and specificity
| Panels | AUC, % | Sensitivity, % | Specificity, % | Accuracy, % |
| RUSC-AS1, MYO15B, ANKRD46 | 91 | 89 | 83 | 85 |
| APCDD1L, RUSC1-AS1 | 88 | 96 | 61 | 77 |
| RUSC1-AS1, MYO15B, MXRA5, ANKRD46 | 88 | 82 | 85 | 84 |
| APCDD1L, RUSC-AS1, MYO15B | 88 | 81 | 81 | 81 |
| APCDD1L, RUSC-AS1, ANKRD46 | 88 | 93 | 69 | 80 |
| RUSC1-AS1, MYO15B, THBS2, ANKRD46 | 87 | 81 | 84 | 82 |
| APCDD1L, RUSC1-AS1, THBS2 | 87 | 89 | 65 | 76 |
| APCDD1L, MYO15B, THBS2 | 87 | 77 | 88 | 83 |
| RUSC1-AS1, THBS2, MXRA5, ANKRD46 | 86 | 82 | 82 | 82 |
| RUSC1-AS1, EXOC2 | 86 | 76 | 88 | 83 |
Conclusions
Our approach shows promising results for designing multiplex MSRE-qPCR panels to accurately predict TNBC response to NAC. Further verification of its efficacy is required on validation cohorts.
Clinical trial identification
Editorial acknowledgement
Legal entity responsible for the study
Research Centre for Medical Genetics, Moscow, Russia.
Funding
The research was supported by Russian Science Foundation (project Nº 18-15-00430).
Disclosure
All authors have declared no conflicts of interest.