Understanding the mechanisms of infectious diseases involves comprehensively examining proteins, including their interactions with other biomolecules. Studies have demonstrated limitations in experimental methods for identifying Protein-Protein Interactions (PPIs), leading to the widespread adoption of Machine Learning (ML) methods for predicting potential PPI candidates, which are often crucial in addressing infectious diseases. More recently in the ML domain, the adoption of the Deep Forest (DF) method for predicting potential PPI candidates has emerged. As traditional or enhanced DF techniques become increasingly widespread in PPI analysis, there arises an urgent need for a review that comprehensively evaluates and scrutinizes these innovative advancements. In response to this need, this paper identifies and reviews PPI prediction models based on the Deep Forest framework, a model suitable for both small and large datasets. This review delves into the unique features, contributions, and limitations of these models, as well as discusses patterns and potential areas for improvement. Additionally, a performance evaluation of the identified Deep Forest models was conducted, with an accuracy ranging between 0.8877 and 0.9930. In conclusion, advanced feature extraction methods and optimization techniques were predominantly the areas of contribution observed in the Deep Forest, as evidenced by the reviewed studies.