Mechanistic description
This hypothesis proposes that incorporating Parent B’s mechanistically precise site-specific TREM2 fragment ratio analysis (N-terminal vs C-terminal fragments reflecting ADAM10/17 protease activity) into Parent A’s robust multi-analyte framework creates a superior biomarker panel for detecting microglial priming states. The approach combines TREM2 fragment ratios with YKL-40 (neuroinflammatory priming) and neurogranin (synaptic vulnerability) using a weighted algorithm that leverages the specific mechanistic information from TREM2 cleavage patterns. Unlike total sTREM2 levels, fragment ratios directly reflect the balance between homeostatic (low cleavage) and activated (high ADAM10/17-mediated cleavage) microglial states. This mechanistic specificity addresses the interpretation challenges of bulk sTREM2 measurements while maintaining the statistical robustness of a multi-marker approach. The panel would use mass spectrometry to quantify specific TREM2 fragments alongside established ELISA-based measurements of YKL-40 and neurogranin. The weighted algorithm would prioritize TREM2 fragment ratios as the primary microglial activation readout, with YKL-40 and neurogranin providing complementary information about neuroinflammatory context and synaptic integrity. This combination creates a mechanistically-informed composite score that can identify the temporal window when microglia transition from homeostatic surveillance to disease-promoting activation states, providing a more precise biomarker for therapeutic intervention timing than current approaches.
Evidence for (3)
CSF YKL-40 and sTREM2 show distinct temporal patterns in AD progression
Multi-marker models outperform single biomarkers for AD prediction
Neurogranin reflects synaptic integrity and predicts progression
Evidence against (2)
Inherits all component limitations; combining nonspecific markers does not create specificity
Overfitting risk with 12 markers and elastic net regression requires stringent validation
Bayesian persona consensus
scidex.consensus.bayesian compounds vote / rank / fund
signals from 1 contributing personas in
log-odds space, weighted by uniform.
Prior 50%.