Alzheimer's Disease (AD) is one of the most common forms of dementia and neurodegenerative diseases, characterised by the formation of neuritic plaques and neurofibrillary tangles. Many different proteins participate in this complicated pathogenic mechanism, and missense mutations can alter the functions of these proteins, significantly increasing the risk of AD. Here we present a machine learningbased analysis to classify the AD-causing mutations from their benign counterparts in 21 AD-related proteins. We first built a generic predictive model having a blind test Matthew’s Correlation Coefficient (MCC) up to 0.74. Performance was further improved at the protein-specific level by tuning the sample weights in the training process (MCC = 0.79). Further to that, our protein-specific model presented the best performance on an independent clinical validation test set, when compared to state-of-the-art methods, suggesting potential clinical utility. Qualitative comparison and feature interpretation indicated that the change of protein stability, hydrophobic environment, and polar interaction contacts were crucial to the pathogenicity of missense mutations. Finally, we developed a user-friendly web server, AlzDiscovery, for researchers to browse the predicted phenotypes of all possible missense mutations on these 21 AD-related proteins. Our study will be a valuable resource for AD screening and the development of personalised treatment.