Una encuesta unificada sobre la predicción de la interacción fármaco-diana y la afinidad de unión: Modelos,...

Área de la Ciencia:

• Computational chemistry and cheminformatics
• Bioinformatics and computational biology
• Drug discovery and development

Sus antecedentes:

• Drug discovery requires accurate prediction of drug-target interactions (DTI) and binding affinity (DTA).
• Experimental methods for DTI and DTA prediction are costly, slow, and have high failure rates.
• Computational prediction methods offer a cost-effective and efficient alternative, improving drug candidate screening.

Objetivo del estudio:

• To systematically review the research progress in computational drug-target interaction (DTI) and drug-target binding affinity (DTA) prediction.
• To highlight the emergence and significance of large pre-trained models as a distinct paradigm in DTI/DTA research.
• To provide a comprehensive overview of data resources, computational models, and experimental considerations in the field.

Principales métodos:

• Review of existing literature on computational methods for DTI and DTA prediction.
• Categorization of methods based on data representation, model evolution (from feature-driven to pre-trained models), and experimental validation.
• Analysis of commonly used resources, evaluation metrics, cold-start problem handling, and model interpretability.

Principales resultados:

• Detailed outline of the technological evolution of computational models for DTI/DTA prediction, including sequence-based, structure-based, and large pre-trained models.
• Discussion of data and representation strategies, computational problem definitions, and experimental best practices.
• Identification of key challenges and future research directions in the field.

Conclusiones:

• Computational methods, especially large pre-trained models, are crucial for efficient and successful drug discovery.
• A comprehensive understanding of model evolution, data handling, and experimental validation is essential for advancing DTI/DTA prediction.
• Future research should focus on addressing current challenges and exploring novel approaches to further enhance prediction accuracy and applicability.