The human hippocampus can pattern separate memories by meaning.

Polypharmacology is an emerging strategy of design, development, and clinical implementation of multi-target-directed ligands (MTDLs) i.e., agents capable of modulating two or more molecular targets simultaneously [, ]. This approach, which emerged beyond the classical ‘one drug-one target’ paradigm, is particularly relevant in the context of complex diseases, such as cancer, neurodegenerative disorders, inflammatory and cardiovascular diseases, where the involvement of multifaceted etiological components limits the effectiveness of single-target-based therapies [, ]. The demand for such innovative medications is increasing also due to limits in clinical efficacy, the occurrence of adverse drug reactions (ADRs), and multidrug resistance, especially seen in the context of infectious diseases [] and cancer [, , ]. In guiding multi-target drug design, the integration of data, clustered regularly interspaced short palindromic repeats (CRISPR) functional screens, and pathway simulations plays a pivotal role in target selection and multi-target ligand prediction []. Assumptions of network pharmacology enable the construction of drug-target networks []. Artificial intelligence (AI) algorithms offer deep generative models along with scalable and versatile platforms able to identify potential drug targets, predict the efficacy, and optimize lead compounds [, ]. Along with the approach of machine learning, they facilitate the generation of molecules with the most optimal balance between potency, selectivity, and safety []. For example, at the early stage of the development of kinase inhibitor candidates, selectivity profiles are typically obtained, and non-kinase inhibitory activity is predicted []. However, the dynamic, self-learning nature of AI systems needs full regulatory guidance, continuous validation, and ethical oversight to enable trustworthy AI deployment in the design of next-generation multi-target compounds []. On the one hand, polypharmacology offers the chance to reduce ADRs and improve patient compliance in comparison to ‘classic’ combination therapies, based on highly selective ligands [, ]. This can be achieved not only due to a generally more predictable pharmacokinetic profile of MTDLs than of drug combinations, but also due to the advantage of using a single formulation and simplification of dosing regimens [, ]. Additionally, complementary synergistic effects may permit the desired therapeutic outcome to be achieved at lower doses of multi-target agents, potentially minimizing the risk of ADRs. On the other hand, engagement of MTDLs raises safety concerns. , an inherent consequence of the promiscuous nature of those molecules, might contribute to adverse reactions that are not fully predictable from preclinical in vitro and in vivo models or early-phase clinical trials []. Such effects may arise from the convergence of pharmacodynamic interactions or tissue-specific target expression, particularly in (highly) heterogeneous patient populations. These challenges emphasize the critical importance of post-marketing drug safety surveillance, as even comprehensive pre-approval evaluations may fail to identify rare, delayed, or population-specific ADRs []. Moreover, fast-track approvals, especially in the case of drugs with orphan designation (see glossary in Table ), are associated with more unreported ADRs, along with more frequent post-marketing safety updates []. Real-world pharmacovigilance systems – incorporating spontaneous reporting databases, electronic health records, and large-scale observational studies – are therefore essential to detect safety signals that emerge only after widespread clinical exposure [].