Journal: bioRxiv

Article Title: SynGlue: AI-Driven Designer for Clinically Actionable Multi-Target Therapeutics

doi: 10.1101/2025.08.28.672835

Figure Lengend Snippet: (a) SynGlue’s integrative workflows for de novo design of PROTACs and multitarget molecules (MTMs), combining data-driven and structure-guided modules for comprehensive design and linker optimization. (b) A comparative analysis showcases Androgen Receptor (AR)-XIAP (X-linked inhibitor of apoptosis protein) PROTACs designed using SynGlue’s data-driven (left) and structure-guided (right) approaches, with a Principal Component Analysis (PCA) plot below illustrating the enhanced chemical diversity achieved by SynGlue while preserving predicted target relevance. (c) Examples of bifunctional, multi-targeting molecules designed using the SynGlue workflow are provided, including MTMs targeting Epidermal Growth Factor Receptor (EGFR) and Cyclin-Dependent Kinase 1 (CDK1) (linked warheads CHEMBL939 and 4442620) and MTMs designed for Protein Kinase C-related kinase X (PRKX) and Dopamine Receptor D1 (DRD1) (using warheads CHEMBL2152768 and CHEMBL1201356). (d) Similarly, a comparative visualization displays Bromodomain-containing protein 4 (BRD4) PROTACs designed using SynGlue’s data-driven (left) and structure-guided (right) approaches, accompanied by a PCA plot below indicating the greater chemical diversity of the SynGlue-designed AR PROTACs while maintaining predicted target relevance. (e) Multi-stage warhead filtering and prioritization, incorporating BRD analog selection, physicochemical filtering (MW, LogP, TPSA, rotatable bonds, aromatic fraction, SA, Csp³), novelty assessment, and final warhead selection. (f) Distribution of Tanimoto similarity scores for selected novel warheads relative to known BRD family warheads. (g) Dendrogram depicting chemical diversity and clustering of selected warheads, illustrating structural novelty and diversity of the SynGlue-optimized warhead library. (h) A flow diagram further details the molecular docking strategy employed for validating or prioritizing the designed warheads, utilizing the experimentally determined structure of Bromodomain-containing protein 4 (BRD4) (Protein Data Bank identifier: 3MXF). (i) Representative chemical structures of prioritized warheads, highlighting their structural diversity and novelty. A tabular representation provides the chemical structures of SynGlue-designed warheads against the BRD4 protein. (j) Interaction map of an optimized warhead within BRD4’s binding cavity, showing key protein-ligand interactions. (k) Predictive modeling workflow incorporating component-wise and linker-specific features, with models trained for DC 50 and D max regression as well as linker-specific multi-class classification using scikit-learn (v1.3.0) and cross-validation. (l) Performance metrics for the predictive models (R², MAE, MSE, RMSE) and density plots for predicted D max and log-transformed DC 50 distributions, confirming model robustness. (m) The chemical structures of two BRD4-targeting PROTACs designed using the SynGlue workflow are shown, accompanied by a table detailing comparative physicochemical profiles and predicted biological activity metrics of these two selected BRD4-specific SynGlue-designed PROTACs.

Article Snippet: Membranes were probed with primary antibodies specific for BRD2 (Cell Signaling Technology [CST], #5848), BRD3 (Santa Cruz Biotechnology, #sc81202), BRD4 (CST, #13440), and c-Myc (CST, #5605). β-Actin (CST, #3700S) or β-Tubulin (CST, #86298) were used as loading controls.

Techniques: Preserving, Selection, Binding Assay, Biomarker Discovery, Transformation Assay, Activity Assay

Journal: bioRxiv

Article Title: SynGlue: AI-Driven Designer for Clinically Actionable Multi-Target Therapeutics

doi: 10.1101/2025.08.28.672835

Figure Lengend Snippet: (a) A bar plot displays the Growth Inhibition 50 (GI 50 ) values in nanomolar (nM) for SynGlue-designed PROTACs, Compound 1 and Compound 2, as determined by CellTiter-Glo® assay across various cancer cell lines, including MV-4-11, VCaP, LNCaP, and MCF7. (b) A schematic representation outlines the experimental strategy employed to assess the in vitro Bromodomain-containing protein 4 (BRD4) protein degradation potential of the SynGlue-designed PROTACs in VCaP, LNCaP, and HeLa cell lines following compound treatment for a specified duration (e.g., 6 hours). (c) Western blots illustrating the dose-dependent degradation of BRD4, BRD2, and BRD3 proteins in VCaP, LNCaP, and HeLa cell lines treated with increasing concentrations (0.1 nM to 100 nM) of Compound 1 and Compound 2 over 6 hours, with β-Actin or β-Tubulin serving as loading controls; the selective BRD4 degradation by Compound 1 is particularly evident in HeLa cells. (d) A Western blot demonstrating the proteasomal degradation of Compound 1-mediated BRD4 degradation, showing that pre-treatment of HeLa cells with a proteasome inhibitor (e.g., MG132 or Bortezomib) followed by Compound 1 treatment rescues BRD4 protein levels compared to treatment with Compound 1 alone. (e) A Western blot illustrates the requirement for a functional PROTAC structure for degradation, as treatment of cells (e.g., HeLa) with an inactive isomer of Compound 1 (AU-16914) at equivalent concentrations fails to induce BRD4 degradation compared to the active Compound 1. (f) Schematic of the in vivo pharmacokinetic (PK) study design, outlining single intravenous (IV) administration of Compound 1 in male CD-1 mice at 5, 10, and 15 mg/kg, with the corresponding plasma concentration time profiles (ng/mL versus hours) shown in line plots to determine PK parameters (C max , AUC, half-life). (g) Line plots showing the maximum tolerated doses (MTD) and PK profiles of Compound 1, with consistent body weight maintenance across 2.5, 5, and 10 mg/kg/day IV doses over the 14-day administration period, indicating good tolerability. (h) Tolerability assessment of Compound 1 in vivo, with line plots tracking average body weight (% change) of male CD-1 mice during the administration and treatment periods, confirming the absence of significant systemic toxicity at tested doses. (i) A schematic representation outlines the in vivo anti-tumor efficacy study design using a relevant xenograft model (e.g., MV-4-11 Cell Line Derived Xenograft (CDX) in athymic nude mice), including treatment groups (vehicle control, Compound 1 at 10 mg/kg daily or alternate day IV, treatment duration (e.g., 17 days), and primary endpoints (tumor volume, body weight). (j) Line plots tracking tumor volume (mm³, mean ± SEM) in treated mice, demonstrating robust tumor growth inhibition by Compound 1 relative to vehicle control, with dose-dependent efficacy. (k) A scatterplot summarizes the effect of Compound 1 treatment on tumor volume (cubic millimeters (mm³), Mean ± Standard Error of the Mean (SEM)) at the study endpoint in the xenograft model, comparing vehicle control, Compound 1 treatment groups, with statistical significance indicated. (k) Line plots of average body weight changes during the in vivo efficacy study, confirming tolerability across all treatment groups. Experimental setup for evaluating the efficacy of compound 1 in mice: Male athymic nude mice (7 to 8 weeks old, 10 mice per group) were inoculated with MV4-11 cells (ATCC) at a density of 15 million cells per mouse in 200□μ□ (1:1 HBSS & ECM gel). Treatment was initiated 8 days after inoculation, when the mean tumor volume reached approximately 170□mm³, and continued for 17 days. Statistical analysis was performed using GraphPad Prism (version 10.2.3), applying Brown-Forsythe and Welch ANOVA tests. (m) Table summarizing the measured concentrations of Compound 1 in plasma and tumor tissues at selected post-treatment time points in the efficacy study, demonstrating sustained intratumoral exposure and a favorable pharmacokinetic–pharmacodynamic (PK-PD) correlation. Compound 1 exposure was assessed in plasma samples (n□=□9-10/group) and tumor tissues (n□=□8-10/group) at the 1-hour time point following the end of the efficacy study. (n) Representative Western blots show the protein levels of BRD4, BRD2, BRD3, and the downstream target c-Myc in tumor tissue lysates harvested from mice at baseline (pre-treatment) and at early time points (1 and 4 hours) post-treatment with Compound 1, demonstrating pharmacodynamic target engagement and pathway modulation. Representative Western blots show the protein levels of BRD4, BRD2, BRD3, and c-Myc in tumor tissue lysates harvested from mice at later time points (6, 8 and 24 hours) post-treatment with Compound 1, illustrating the duration of pharmacodynamic effects.

Article Snippet: Membranes were probed with primary antibodies specific for BRD2 (Cell Signaling Technology [CST], #5848), BRD3 (Santa Cruz Biotechnology, #sc81202), BRD4 (CST, #13440), and c-Myc (CST, #5605). β-Actin (CST, #3700S) or β-Tubulin (CST, #86298) were used as loading controls.

Techniques: Inhibition, Glo Assay, In Vitro, Western Blot, Functional Assay, In Vivo, Clinical Proteomics, Concentration Assay, Derivative Assay, Control, Drug discovery