Figure 1.
Calculation of Matthews Correlation Coefficient (MCC).
(A) Contingency matrix illustrating our usage of true negatives (TN), false positives (FP), false negatives (FN) and true positives (TP). (B) Mathematical definition of the MCC.
Table 1.
Baseline demographics.
Figure 2.
Detectability in the ADNI dataset of the 18 proteins highlighted by Ray et al. (2007).
Of the 18 proteins in the signature highlighted by Ray and colleagues [9], three were below the detection limits of the ADNI assay, 11 were considered detectable by ADNI and four were not assessed. Protein abbreviations are defined in Table S2.
Table 2.
Accuracy of the analytes that passed entropy filtering in classifying control and MCI progressor samples.
Figure 3.
APOE plasma concentration as a function of APOE genotype.
APOE concentration shows a decreasing trend from ‘protective’ (ε2) to ‘at risk’ (ε4) genotypes that is independent of clinical diagnosis. Plots illustrate APOE log10 plasma concentrations as a function of APOE genotype when considering samples classified at baseline as (A) control (n = 54), (B) MCI (n = 396) and (C) AD (n = 112). Plot (D) illustrates the trend when baseline diagnosis is not considered. 2/3 – APOE-ε2/ε3; 2/4 – APOE-ε2/ε4; 3/3 - APOE-ε3/ε3; 3/4 - APOE-ε3/ε4; 4/4 - APOE-ε4/ε4. Statistically significant (p<0.05) difference when compared to a2/3, b2/4, c3/3, d3/4. Statistical tests were not conducted on sample sets of n<4.
Table 3.
Feature set selection of 11-analyte signatures to discriminate control and MCI progressor samples.
Table 4.
Accuracy of analyte signatures in classifying controls and MCI progressors.
Table 5.
Minimal meta-feature set selection of differences in analyte abundances to discriminate Control and MCI Progressor samples.
Table 6.
Accuracy of meta-feature signatures involving differences in analyte abundances in classifying controls and MCI progressors.
Table 7.
Accuracy of the analytes that passed entropy filtering in classifying control and AD samples.
Table 8.
Validation of proposed signatures on data collected at 12 month follow-up.
Table 9.
Feature set selection of signatures to discriminate control and MCI progressor samples based on longitudinal change.