Table 7. Eliminating the poorest quality samples.
Example of a strategy that used all three sample cut-points and eliminated the poorest quality samples and chips. The criteria applied for inclusion were: AFS = 0 AND pH ≥ 6.6 AND 28S/18S ≥ 1.5. The samples that passed all three measures were placed into the ‘included’ group and the remaining were placed in the group ‘excluded.’ The two groups were compared by t-test separately in two brain regions. The significant p-values are in bold. The abbreviations are the same as Table 1. The microarray chip quality was improved in the included compared to the excluded samples.
| Clinical Quality | Tissue Quality | RNA Quality | Microarray Chip Quality | |||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
|
| ||||||||||||||
| Brain | Criteria | Age | AFS 0 / ≥1 |
pH | PMI (hours) |
Freezer Time (years) |
28S/18S | 3′/5′ GAPDH | 3′/5′ ACTB | RIN** | %PC | SF | ACI | Type 1/2 1 / 2*** |
| ACC | Include (n = 60) | 49.43 (13.76) | N/A | 6.93 (0.16) | 22.96 (7.79) | 3.35 (3.36) | 1.98 (0.22) | 1.46 (0.37) | 2.49 (1.42) | 5.94 (0.93) | 44.00 (4.23) | 1.07 (1.03) | 0.97 (0.02) | 56 / 4 |
| Exclude (n = 38) | 56.26 (12.05) | N/A | 6.66 (0.26) | 25.52 (9.19) | 2.51 (3.41) | 1.51 (0.36) | 2.03 (0.93) | 3.60 (2.13) | 5.67 (1.58) | 41.47 (6.25) | 1.97 (2.08) | 0.92 (0.05) | 12 / 24 | |
| p-value | 0.01 | N/A | N/A | 0.16 | 0.24 | N/A | 0.00074 | 0.0006 | 0.35 | 0.03 | 0.02 | <0.000005 | <0.000005 | |
|
|
||||||||||||||
| CB | Include (n = 59) | 50.20 (13.81) | N/A | 6.94 (0.16) | 23.44 (7.99) | 3.11 (3.24) | 2.06 (0.34) | 1.42 (0.48) | 3.02 (2.74) | 7.11 (1.31) | 45.96 (3.08) | 1.66 (3.10) | 0.98 (0.02) | 57 / 2 |
| Exclude (n = 32) | 54.31 (11.75) | N/A | 6.62 (0.25) | 24.25 (9.26) | 3.22 (3.73) | 1.57 (0.50) | 2.16 (0.97) | 4.54 (3.33) | 5.72 (1.81) | 40.20 (6.86) | 4.24 (5.28) | 0.92 (0.07) | 7 / 25 | |
| p-value | 0.14 | N/A | N/A | 0.68 | 0.89 | N/A | 0.0002 | 0.03 | 0.002 | 0.00006 | 0.02 | 0.0005 | <0.000005 | |
N/A. The a priori selection of quality by AFS, pH, and RNA produced highly significant differences in these 3 measures; therefore the p-value was not shown.
RIN calculation for ACC only included part of the sample due to technical difficulties in the Agilent run.
Type 1 / 2 p-values were calculated by Fisher's exact test.