Abstract
The growing availability of structured data types, including molecular and pharmacological data, along with unstructured data types such as medical imaging data, has enabled the development of statistical, machine learning (ML), and deep learning (DL) approaches for drug response prediction. These computational methods are integral to precision medicine, leveraging data-driven techniques to predict patient-specific treatment outcomes. This review provides a systematic overview of existing methodologies for drug response prediction, focusing on input data structures, response variable definitions, and data types utilized. In contrast to previous reviews that focus on specific therapies or computational approaches, we present a unified classification framework based on data-response relationships, including single data type with a response vector, single data type with a response matrix, and multiple data types with a response. By using this structure, we can compare statistical and ML-based models across different diseases and data types. Finally, we discuss evaluation strategies, highlight emerging methodological trends, and outline key challenges and future opportunities to advance drug response prediction.