Figure 1.

An overview of the approach. (A) The common strategy today is to look for novel biomarkers that would classify Alzheimer’s disease (AD) patients in a more accurate, patient-specific manner. In the illustrated example, four AD biomarkers were tested (genes A–D). Based on the expression levels of those markers (B) the patients are classified. Patients 1 and 2 in this example have similar expression levels of biomarkers A and B, and therefore would be classified as molecularly similar (C). We explore AD pathology in an unbiased manner (D–F). The workflow of our approach consists of patient-specific “omic” profiling, followed by surprisal analysis (D), aiming to decipher not only altered transcripts/proteins, but also the structure of the altered network, namely the patient-specific altered transcriptional signature (E,F). This signature is composed of distinct unbalanced processes, each resulting from a constraint that operates on the system (see main text). Overexpression of the biomarker B in patient 1 is associated with the black and green processes, whereas in patient 2 it is upregulated due to only the black process (D–F).