Unlocking the potential of marine natural product fragments for rational anticancer drug design: a computational...

Marine natural products (MNPs) offer a vast and diverse reservoir of chemically rich compounds that remain underexplored for fragment-based drug discovery (FBDD) against cancer. In this study, we developed Marine-FL, a fragment library derived from the Comprehensive Marine Natural Products Database (CMNPD) and applied it within an in silico FBDD pipeline targeting four proteins, CLU, PD-1, PD-L1, and CTLA-4, which are implicated in chemoresistance and immune evasion. CMNPD molecules were standardized, filtered using Rule-of-Three criteria, and systematically fragmented with the BRICS algorithm, followed by deduplication and capping to generate a 4,643-fragment library for chemical space analysis. The library was screened using AutoDock Vina and rescored with the neural network-based AKScore2 model to prioritize fragments for each receptor for the subsequent steps. Top candidates were further optimized by DeepFrag-guided fragment growing, and a representative complex underwent explicit-solvent MD simulations and MM/PB(GB)SA binding free energy calculations to assess stability and relative affinity. PCA and UMAP analyses showed that the fragment library spans broad and sparsely clustered regions, with Murcko scaffold diversity metrics indicating both recurrent cores and a high proportion of rare singletons, indicating a balance between familiar chemotypes and a considerable reserve of structurally novel fragments. Notably, certain multi-ring, conjugated scaffolds recurred among computationally top-ranked fragments across the four target proteins, revealing structural motifs that merit experimental investigation for polypharmacology-directed validation. MD simulations and binding energy calculations of Ligand 10 with PD-L1 provided proof-of-concept dynamic stability at triplicate 200-ns runs and favorable binding energetics with MMPBSA mean delta ΔG = - 8.1 to - 5.6 kcal/mol. Exploratory single-trajectory simulations with PD-1, CTLA-4, and sCLU showed stable complex formation, supporting computational multi-target compatibility pending experimental validation. These findings support the utility of diverse marine-derived fragment libraries as sources of scaffolds with broad applicability, advancing sustainable, structure-guided approaches in anticancer FBDD.