Quantitative Structure Activity Relationship (QSAR) modeling is a constructive methodology of drug discovery, which allows forecasting the biological activity based on the chemical structure. However, traditional QSAR techniques usually find it difficult to encompass complex non-linear relationships in large-scale molecular data. Recent developments in artificial intelligence have seen the development of Deep QSAR, which combines the use of deep learning architectures including convolutional and recurrent neural networks and molecular representations. The technique enhances forecasting efficacy, toxicity and safety profiles of drugs. Deep QSAR models are superior to traditional algorithms, especially in high-dimensional chemical data, in that they can learn the appropriate features without a lot of manual engineering of descriptors. Moreover, this is combined with generative models which enable the creation of novel compounds with idealized biological characteristics. With the further development of computational power and data availability, Deep QSAR is projected to become an important factor in faster drug development, cost-saving and more successful translational outcomes. The problem of model interpretability and data quality is still a persistent issue in future research.

Historically, animal testing has been the main focus in drug development but ethical issues, high expenses and low translational applicability have led to seeking alternative methods. The global ethics theory or the 3Rs framework of Reduce, Refine and Replace, is an ethical framework used to ensure that the use of animals is reduced, but at the same time, can be scientifically valid. The development of in vitro systems, organ-on-chip models, computational models and artificial intelligence has contributed significantly to the empowerment of non-animal research plans. These inventions allow better forecasting of the drug efficacy and toxicity and minimizing animal numbers and increase their welfare when the use of animals is still required. Artificial intelligence also supplies the 3Rs with better experimental design, better interpretation of data and real-time welfare monitoring. Collectively, these strategies enhance more effective, humane and predictive drug discovery. The further incorporation of the new technologies in the 3Rs framework will result in the improvement of population health and the alignment of pharmaceutical research to the ethical and regulatory demands.

Background: Proliferating cell nuclear antigen (PCNA) is a major controller of DNA replication and repair and is vital in the survival of cancerous cells. Cancer-specific isoform of PCNA (caPCNA) displays structural divergence with normal PCNA, providing a target of treatment. The first-in-class small-molecule inhibitor, AOH1160 is developed to take advantage of this specificity.

Methods: The systematic literature review was performed on the leading scientific databases to find preclinical trials assessing AOH1160. Relevant studies were filtered according to their eligibility criteria. The studies were evaluated based on molecular targeting, anticancer activity, pharmacokinetics, safety, and potential of combination therapy.

Results: The studies that have been included consistently show that AOH1160 selectively binds the L126-Y 133 region of caPCNA, which inhibits the DNA replication and homologous recombination repair of cancer cells. Treatment caused cell-cycle arrest, apoptosis and damage to DNA without damaging normal cells. In vivo studies justified tumour suppression, oral bioavailability, desirable pharmacokinetics and low toxicity. The cisplatin synergistic effect was also reported.

Conclusion: The preclinical evidence of AOH1160 is good as a selective anticancer agent against caPCNA. To ensure safety and therapeutic efficacy, its clinical validation is necessary.