RESUMO
Deep convolutional neural networks (CNNs) are a method of choice for image recognition. Herein a hybrid CNN approach is presented for molecular pattern recognition in drug discovery. Using self-organizing map images of molecular pharmacophores as input, CNN models were trained to identify chemokine receptor CXCR4 modulators with high accuracy. This machine learning classifier identified first-in-class synthetic CXCR4 full agonists. The receptor-activating effects were confirmed by intracellular cAMP response and in a phenotypic spheroid invasion assay of medulloblastoma cell invasion. Additional macromolecular targets of the small molecules were predicted inâ silico and tested inâ vitro, revealing modulatory effects on dopamine receptors and CCR1. These results positively advocate the applicability of molecular image recognition by CNNs to ligand-based virtual compound screening, and demonstrate the complementarity of machine intelligence and human expert knowledge.
Assuntos
Movimento Celular , Aprendizado Profundo , Receptores CXCR4/agonistas , Receptores CXCR4/antagonistas & inibidores , Linhagem Celular Tumoral , Desenho de Fármacos , HumanosRESUMO
PURPOSE: Aberrant activation of the fibroblast growth factor receptor (FGFR) family of receptor tyrosine kinases drives oncogenic signaling through its proximal adaptor protein FRS2. Precise disruption of this disease-causing signal transmission in metastatic cancers could stall tumor growth and progression. The purpose of this study was to identify a small molecule ligand of FRS2 to interrupt oncogenic signal transmission from activated FGFRs. METHODS: We used pharmacophore-based computational screening to identify potential small molecule ligands of the PTB domain of FRS2, which couples FRS2 to FGFRs. We confirmed PTB domain binding of molecules identified with biophysical binding assays and validated compound activity in cell-based functional assays in vitro and in an ovarian cancer model in vivo. We used thermal proteome profiling to identify potential off-targets of the lead compound. RESULTS: We describe a small molecule ligand of the PTB domain of FRS2 that prevents FRS2 activation and interrupts FGFR signaling. This PTB-domain ligand displays on-target activity in cells and stalls FGFR-dependent matrix invasion in various cancer models. The small molecule ligand is detectable in the serum of mice at the effective concentration for prolonged time and reduces growth of the ovarian cancer model in vivo. Using thermal proteome profiling, we furthermore identified potential off-targets of the lead compound that will guide further compound refinement and drug development. CONCLUSIONS: Our results illustrate a phenotype-guided drug discovery strategy that identified a novel mechanism to repress FGFR-driven invasiveness and growth in human cancers. The here identified bioactive leads targeting FGF signaling and cell dissemination provide a novel structural basis for further development as a tumor agnostic strategy to repress FGFR- and FRS2-driven tumors.
Assuntos
Descoberta de Drogas , Neoplasias Ovarianas , Animais , Feminino , Humanos , Camundongos , Proteínas Adaptadoras de Transdução de Sinal/química , Proteínas Adaptadoras de Transdução de Sinal/metabolismo , Ligantes , Proteínas de Membrana/química , Proteínas de Membrana/metabolismo , Neoplasias Ovarianas/tratamento farmacológico , Proteoma/metabolismo , Receptores de Fatores de Crescimento de Fibroblastos/metabolismo , Transdução de Sinais/fisiologia , Descoberta de Drogas/métodosRESUMO
Constructive machine learning aims to create examples from its learned domain which are likely to exhibit similar properties. Here, a recurrent neural network was trained with the chemical structures of known cell-migration modulators. This machine learning model was used to generate new molecules that mimic the training compounds. Two top-scoring designs were synthesized, and tested for functional activity in a phenotypic spheroid cell migration assay. These computationally generated small molecules significantly increased the migration of medulloblastoma cells. The results further corroborate the applicability of constructive machine learning to the de novo design of druglike molecules with desired properties.