This study significantly expands the structural coverage of the human proteome by applying the state-of-the-art machine learning method, AlphaFold, to predict structures for almost the entire human proteome (98.5% of human proteins). The resulting dataset covers 58% of residues with a confident prediction, and 36% of residues have very high confidence. The authors introduce several metrics, including pLDDT and pTM, to interpret the dataset and identify multi-domain predictions and disordered regions. Case studies on glucose-6-phosphatase, diacylglycerol O-acyltransferase 2, and wolframin highlight how high-quality predictions can generate biological hypotheses. The predictions are freely available, and the authors anticipate that routine large-scale and high-accuracy structure prediction will become an important tool for addressing new questions from a structural perspective.This study significantly expands the structural coverage of the human proteome by applying the state-of-the-art machine learning method, AlphaFold, to predict structures for almost the entire human proteome (98.5% of human proteins). The resulting dataset covers 58% of residues with a confident prediction, and 36% of residues have very high confidence. The authors introduce several metrics, including pLDDT and pTM, to interpret the dataset and identify multi-domain predictions and disordered regions. Case studies on glucose-6-phosphatase, diacylglycerol O-acyltransferase 2, and wolframin highlight how high-quality predictions can generate biological hypotheses. The predictions are freely available, and the authors anticipate that routine large-scale and high-accuracy structure prediction will become an important tool for addressing new questions from a structural perspective.