摘要： Alzheimer’s disease (AD) is one of the major global health challenges of the 21st century. More than 200 distinct mutations in presenilin 1 (PSEN1) cause severe early-onset familial AD (FAD) and are thus of central interest to the etiology of AD. PSEN1 is the catalytic subunit of γ-secretase that produces β-amyloid peptide (Aβ), and the mutations tend to increase the produced Aβ42/Aβ40 ratio. The molecular reasons for the pathogenesis of these mutations are unknown. We studied a close-to-complete data set of PSEN1 mutations using 21 different computational methods hypothesized to reproduce pathogenesis, using both sequence- and structure-based methods with the full γ-secretase complex as input. First, we tested whether pathogenicity can be estimated accurately using all possible mutations in PSEN1 as a direct control. Several methods predict the pathogenicity of the mutations (pathogenic vs all other possible mutations) well, with accuracies approaching 90%. We then designed a stricter test for predicting the severity of the mutations estimated by the average clinical age of symptom onset for mutation carriers. Surprisingly, we can predict the clinical age of symptom onset at 95% confidence or higher with several methods. Accordingly, our results show that simple biochemical properties of the amino acid changes rationalize an important part of the pathogenicity of FAD-causing PSEN1 mutations. Although pathogenic mutations generally destabilize γ-secretase, all of the tested protein stability methods failed to predict pathogenicity. Thus, either the static cryogenic-electron-microscopy-derived molecular-dynamics-equilibrated structures used as input fail to capture the stability effect of mutated side chains or protein stability is simply not a key factor in the pathogenicity. Our findings suggest that the chemical causes of FAD may be modeled and lend promise to the development of a semiquantitative model predicting the age of onset of mutation carriers that could eventually become of care-strategic value.