Table 2. . The number of CpG sites showed association with sample physical positions and batches in studies.† .
| Studies (n probes) | Methyl450 datasets | Methyl27 datasets | EPIC datasets | |||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| ROSMAP (161,862) | GSE58885 (156,108) | GSE74193 (149,069) | GSE38873 (10,640) | GSE26133 (11,500) | BrainCloud (11,500) | GSE86831 (724,466) | GSE93373 (724,466) | |||||||||
| Process | Position | Batch | Position | Batch | Position | Batch | Position | Batch | Position | Batch | Position | Batch | Position | Batch | Position | Batch |
| Raw | 52,988 | 152,977 | 46,754 | 8,504 | 68,079 | 126,766 | 0 | 7471 | 1 | 10,301 | 1604 | 7848 | 269,705 | 128,435 | 0 | 0 |
| B(C) | 61,725 | 0 | 52,071 | 1 | 86,622 | 0 | 52 | 29 | 4182 | 0 | 3952 | 0 | 0 | 0 | 6 | 0 |
| P(C) | 112 | 153,775 | 0 | 11,850 | 5 | 127,739 | 0 | 7586 | 0 | 10,433 | 0 | 7934 | 325 | 0 | 0 | 8469 |
| BP(C) | 94 | 0 | 0 | 1 | 21 | 0 | 0 | 35 | 0 | 10,438‡ | 0 | 7956‡ | 0 | 0 | 0 | 14 |
| PB(C) | 137 | 0 | 0 | 2 | 45 | 0 | 0 | 14 | 0 | 0‡ | 6 | 0‡ | 0 | 0 | 0 | 4 |
| P(lm) | 0 | 153,245 | 0 | 12,982 | 0 | 126,546 | 0 | 7599 | 0 | 10,370 | 0 | 7909 | 0 | 0 | 0 | 0 |
| B(C)P(lm) | 0 | 0 | 0 | 1 | 0 | 0 | 0 | 36 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 |
| P(lm)B(C) | 0 | 0 | 0 | 1 | 11 | 0 | 0 | 21 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 |
| FN | 108,296 | 134,263 | NA | NA | 79,781 | 131,478 | NA | NA | NA | NA | NA | NA | 169,372 | 93,441 | 0 | 0 |
| ssNoob | 85,114 | 153,621 | NA | NA | 1656 | 124,890 | NA | NA | 3 | 8105 | NA | NA | 224,821 | 106,030 | 0 | 0 |
Methyl450 means Illumina Infinium HumanMethylation450 BeadChip Array and Methyl27 means Illumina Infinium HumanMethylation27 BeadChip Array. N probes mean the number of the probes after filtering and preprocessing. The result data means the number of loci correlated at methylation sites and position FDR q-value <0.05 in different datasets.
We used ANOVA analysis to calculate the p-values of correlation between methylation levels and position or batch. FDR q-value was computed for each nominal p-value by controlling the FDR at 0.05 using the R function 'qvalue'. We then obtained the number of CpGs significantly associated with positions and batches.
†False discovery rate q-value <0.05.
‡Table cells are notable due to the corrected order.
ANOVA: Analysis of variance; BP: Bipolar; FDR: False discovery rate; FN: Functional normalization; NA: Not available; ssNoob: Single-sample Noob.