The paper introduces the use of amphiphilic pseudo amino acid composition (Am-Pse-AA) to predict enzyme subfamily classes, particularly focusing on oxidoreductases. The Am-Pse-AA composition is a novel representation that combines the conventional amino acid composition with a set of correlation factors reflecting hydrophobicity and hydrophilicity distribution patterns along the protein chain. This approach aims to incorporate sequence-order effects into the prediction process, which were previously overlooked in methods based solely on amino acid composition. The authors developed an augmented covariant-discriminant predictor using the Am-Pse-AA composition. The predictor was tested on a training dataset of 2640 oxidoreductases and an independent dataset of 2124 oxidoreductases. The results showed that the new predictor achieved significantly higher success rates compared to previous methods, both in terms of self-consistency tests and jackknife tests. The optimal value for the number of correlation factors, λ, was found to be 9 for the current dataset. The study also highlights the importance of sequence-order information in predicting enzyme subfamilies, suggesting that different proteins have distinct amphiphilic features corresponding to different hydrophobic and hydrophilic sequence-order patterns. The findings indicate that the Am-Pse-AA composition and the augmented covariant-discriminant algorithm can effectively predict enzyme subfamily classes, providing valuable insights into the molecular mechanisms of enzymes.The paper introduces the use of amphiphilic pseudo amino acid composition (Am-Pse-AA) to predict enzyme subfamily classes, particularly focusing on oxidoreductases. The Am-Pse-AA composition is a novel representation that combines the conventional amino acid composition with a set of correlation factors reflecting hydrophobicity and hydrophilicity distribution patterns along the protein chain. This approach aims to incorporate sequence-order effects into the prediction process, which were previously overlooked in methods based solely on amino acid composition. The authors developed an augmented covariant-discriminant predictor using the Am-Pse-AA composition. The predictor was tested on a training dataset of 2640 oxidoreductases and an independent dataset of 2124 oxidoreductases. The results showed that the new predictor achieved significantly higher success rates compared to previous methods, both in terms of self-consistency tests and jackknife tests. The optimal value for the number of correlation factors, λ, was found to be 9 for the current dataset. The study also highlights the importance of sequence-order information in predicting enzyme subfamilies, suggesting that different proteins have distinct amphiphilic features corresponding to different hydrophobic and hydrophilic sequence-order patterns. The findings indicate that the Am-Pse-AA composition and the augmented covariant-discriminant algorithm can effectively predict enzyme subfamily classes, providing valuable insights into the molecular mechanisms of enzymes.