Fig. 2: Demonstration of useful attributes captured by PCA.
From: AI-QuIC machine learning for automated detection of misfolded proteins in seed amplification assays
a Class Separation Achieved by PCA. A PCA was run on the raw data and metrics data and generated principal components (PCs) corresponding to linear combinations of the input features. The figures demonstrate how these PCs create distinctions in the dataset which can be used with AI. b Edge-case Examples. The PCA plots from a) were used to select the most positive-like false positives for each feature extraction method and compare with clearly defined positive samples. This demonstrates any likenesses which may make it difficult for PCA to create meaningful distinctions.
