* Ml ¦M Integrating i and in vitro data to identify putative thyrotropin- releasing hormone receptor ligands v * ' I Jv* : 't ' ¦ St U j I g m iwjii JP.:; ..' %&. ACS 2020 Fall Meeting Mahmoud Shobair, PhD Post Doctoral Fellow L I. S. Environmental Research Triangle Park, NC % Disclaimer: The views expressed in this presentation are those of the authors and do not necessarily reflect the views or policies of the U.S. EPA. presentation does not represent EPA policy. ------- High-throughput screening (HTS) in risk assessment K-4K cmpds ** ** "*«* ¦ D . "Hi " K« rj N> ToxCast WIWW Tox21 Data Models ------- How many environmentally-relevant chemicals interact with the thyrotropin-releasing hormone receptor (TRHR)? Molecular If a chemical interacts with TRHR, it Experimental approach initiating may disrupt thyroid event (MIE) production. A* Are all hits reliable? Are we missing important hits? How can we increase confidence in results? ------- Tox21_TRHR Assay Design Thyrotropin-releasing BEB1 Hypothesis: Tox21 TRHR assay m the receptor (TRHR) response to its specific ligand (TRH). Steps l-IV can influence hit interpretation ------- Tox21_TRHR Agonist & Antagonist Actives Tox21_TRHR (~ 8000 total screened) 160 antagonist 388 Total ------- Problem Statement & Approach Tox21_TRHR assay provides an indirect measure of potential TRHR activity Large number of diverse environmental chemicals screened, yet false negatives and positives are expected Goal is to identify subset of likeliest true actives from the full set of assay results Approach is to prioritize subset of actives (true hits) and inactives (potential false negatives) for follow-up testing using: > domain knowledge > chemotype enrichments > in silico computational chemistry ------- Examples of most active Tox21_TRHR actives Maximum AUC by chemical is the area under the curve-fit from the ToxCast Pipeline, and compresses potency and efficacy into a single metric. TRH Allyl alcohol Proflavin hydrochloride Digitonin Rhodamine 6G 2-Chloro-1,4-diaminobenzene sulfate Fluorescein 5(6)-isothiocyanate Riboflavin Eosin 2,,4',5',7,-Tetrabromofluorescein 4',5'-Dibromofluorescein Fluorescein sodium Fluorescein ]+ Ctrl Some active substances may be due to assay interference from auto- fluorescence. Fluorescent 0 100 200 300 400 max(AUC) 500 600 700 ------- Privileged scaffolds from chemotype enrichment analysis chain:alkaneLinear_stearyl_C18 group:carbohydrate_hexopyranose_generic atom:element_metal_transistion_metal r i n g: f u sed_ste ro i d_g e n e r i c_[5_6_6_6] group:ligand_path_5_bidentate_propandiamine bond:CS_sulfide_dialkyl bond:CN amine ter-N aromatic o 10 15 20 25 # antagonists 30 ^7""' I" Hg2+1" h3c 35 40 45 phenothiazines ------- Filtering potential sources of false positive hits structure-based Detection technology chemotype enrichment interference heuristic, Non-specific literature support Ca2+ response domain knowledge derived from 2D filters reference db 4 ?? ------- Binding reference dataset and modeling Data cur 10 Journal articles Patents BindingDB Reference dat binding data to: - Eliminate binders wit - Train models 3D pharmacop features r 115 unique ~ structures Data mining r i association rules & activity l values R 3D mode \ pharmacophore models L J (76) Binders rings with > 3 rings (16) Non-binders with > 3 rings (23) Non-binders with < 3 rings (0) Binders with < 3 rings) ------- Binding reference dataset: structural diversity Ki or IC50 0.01 -300 uM 1. Analogs to the natural TRHR ligand bond:C(=0)N_carboxamide_generic (75) 2. Heterocyclic compound ring:hetero_[6]_Z_generic (31) . 4 N* O. o "NH 3. Benzodiazepine-like structures ring:hetero_[6_7]_N_benzodiazepine_(1_4-) (7) h2n^ o o Mil' 11 ------- Filtering potential sources of false positive hits structure-based chemotype enrichment heuristic, literature support 388 hits Detection technology interference Non-specific Ca2+ response domain kno derived reference 2D filters 4 140 filtered hits ~ -8 Hg 2+ ~ -26 HO Y 0H ~ -208 sOH ------- Pharmacophore modeling s: o ts Qj ¦£ to O C2 -Q Qj V) D QJ O ^ ? -C> ^ g ^ ^ 8j Generation of 3D conformations Molecular alignment of binders MOE J r ^ Generation & validation of pharmacophore models l. j C3 O .to D U Qj v -Cl u o Q. Qj ^ to Qj Querying structure database Analysis of model results mol ~F A * & & $¦ A i ti % rmsd 0.4799 0.2784 0.2636 0.5092 0.2334 0.3147 0.4857 0.2842 0.3047 0.3550 0.2474 0.3806 0.3041 0.2422 0.2404 0.2885 0.4383 0.5112 0.2485 0-13482 F5:Hyd F3:Aro|Hyd ------- Prioritizing active hits 388 hits 140 filtered hits MODERATE HIGH [TRH model, benzo model 1, benzo model 2, mixed model 1 , mixed model 2 [0, 1, 0, 1, 0, 1, 1, 0, 1, 0, 0, 0, 0, 0, 1] Concordance: Models Assay & Expert MODERATE: at least one + model quality check 14 Flurazepam F3:Aro|Hyd ------- Predicted TRHR inhibitor: PK-11195 (Moderate binder) 3D Modeling yields plausible results in the dru DTXSID7041097 isoquinoline carboxamide binds selectively to the peripheral benzodiazepine receptor (PBR) Midazolam, known TRHR binder PK-11195 aligned to Midazolam DTXSID8047846 15 ------- Moderate-plausibility assay active: Chlorophacinone Chlorophacinone has structural features associated with Benzodiazepine inhibition ofTRHR [ hit in 5 models] TRHR antagonist DTXSID1023071 DTXSID2032348 ------- High-plausibility assay active TRH F5:Hyd F3:Aro|Hyd Saquinavir mesylate Saquinavir mesylate DTXSID9023835 17 ------- Newly identified TRHR candidate modulators from actives Benzodiazepine Known TRHR competitive (midazolam), assay active /^New atypical Benzodiazepine Opioid antagonist Does not bind GABA receptor, inhibits peptide receptor (CCK) Unclear, further inqu 18 ------- Identifying false negatives 7872 chemicals 2D filter > 2 rings <0 1553 structures for 3D modeling Diverse structures (27) benzodiazepines (3) peptides 19 ------- Prioritized peptide-like structures F5:Hyd F3:Aro|Hyd DTXSID8023551 Assay active DTXSID4047252 DTXSID8046456 ------- Predicted Moderate Binders Structurally diverse Multi-ring structures Primarily drugs O-T^K ~ ~ 0 ~ 0s ~ ~ ~ XX ~ o N s T o ~ ~ ~ / \ ~ ~ O N. NCHi k/ II -/NTi ~ ~ .CH, HC1 ~ '-^ S ~ ~ ~ ~ > - O CI p -N KC ~W ^Gfc o > ' O :r*' r~\ \_/ ~ ~ "V'y; »c [ > V"~f % -~ o a f ~ ~ ~ ~ L XXj ~ u ~ f y f JL ci 21 ------- Summary & Conclusions >50% assay false positive; enriched features associated with artifacts Multistep prioritization workflow combines > existing domain knowledge > in vitro results > in silico computational chemistry Integrated approach points to small number of true active candidates in both the hits and negatives, including a novel benzodiazepine-type structure A limitation of this work is that the 3D modeling assumes the TRHR binding pocket in the native conformation. 3D models predict larger set of structurally diverse moderate binders among hits that are of potential environmental significance, warranting follow-up evaluation ------- Summary & Conclusions Multistep workflow is generalizable and can be applied to other high-throughput assay results to improve ability to filter out false positives and identify potential true actives for follow-up screening. Data Molecular initiating even ^ Adverse ------- Acknowledgements > Chris Grulke > Katie Paul Friedman > Ann Richard > Daniel Chang > Ryan Lougee > Tox21 & ToxCast assay collaborative group ------- |