Table 1.
Demographic information of the studied cohort.
Fig 1.
We detect clusters on the space defined by the blood markers, define a different profile for each cluster, and, using the phenotypes extracted from the MRI images, we analyze the interactions between the profiles and the disease.
Fig 2.
For groups G1 and G2, we create new random subgroups and test them. Then, we correct the obtained result in the original group (Pc, red line), with respect to the 5% percentile of the obtained distribution of p-values (P0.05, black line). If the result is in the percentile, we consider it significant.
Fig 3.
Similarity matrices and 2D embeddings.
a) Euclidean distance between subjects over the initial blood marker space. b) Learnt similarity matrix S. Subjects in the matrix are ordered by the obtained clusters. c) d) 2D embeddings of the respective matrices in a) and b) using t-SNE.
Table 2.
Stability tests.
Table 3.
Demographic characteristics of the different clusters.
Fig 4.
Top 10 marker weights in the kernel combination.
Fig 5.
Distribution of top ten ranked markers for each cluster. Normalized values.
Fig 6.
Differences in volume for each of the presentations against the rest. Corrected using permutation. Inf: Inferior. WM: White Matter. CC: Cingular Cortex.
Fig 7.
Whole cluster analysis, cortical thickness.
Differences in cortical thickness for each of the presentations against the rest. Corrected using permutation.
Fig 8.
Differences between diagnostic groups for each of the presentations. Corrected using permutation. Inf: Inferior. WM: White Matter. CC: Cingular Cortex.
Fig 9.
Diagnostic group analysis, whole population.
Differences between diagnostic groups on the whole population. Inf: Inferior. WM: White Matter. CC: Cingular Cortex.
Fig 10.
Diagnostic group analysis, cortical thickness.
Differences in cortical thickness between diagnostic groups for each of the presentations. Corrected using permutation.
Fig 11.
Diagnosis interaction analysis.
Differences between diagnosis stages across presentations. Corrected using permutation. Inf: Inferior. WM: White Matter. CC: Cingular Cortex.
Fig 12.
Diagnostic interaction analysis, cortical thickness.
Differences in cortical thickness between diagnostic stages across presentations. Corrected using permutation.