Mining ClinicalTrials.gov via CTTI AACT for drug target hypotheses
0 likes88 views
This document discusses methods for mining clinical trial data from ClinicalTrials.gov to infer disease-target associations. It presents several proposed confidence metrics for evaluating these associations and provides examples. Over 1.2 million associations were found involving 164,000 unique disease-gene pairs. Named entity recognition tools were used to extract drugs, diseases, and other information from trial descriptions. Top drugs, diseases, and targets were identified based on number of mentions.
Mining ClinicalTrials.gov via CTTI AACT for drug target hypotheses
- 1. Disease-target associations from drug trial links Proposed confidence metrics: ● nStudy : Study count for association. ● nStudyNewness : Study count weighted by newness of study (newer better). ● nStudyPhase : Study count weighted by phase of study (completed better). ● nPub : Study publications. ● nPubTypes : Study publications (results type better). ● nDiseaseMention : Disease mention count for disease-target association. ● nDrugMention : Drug mention count for disease-target association. ● nDrug : Drug count for disease-target association. ● nAssay : Assay count for drug-target association. ● nAssayPchembl : Assay count for drug-target association, weighted by pChembl. 1.2M associations; 164K unique disease-gene pairs. Examples: Mining ClinicalTrials.gov via CTTI AACT for drug target hypotheses Jeremy Yang1 , Roger Sayle2 , Lars Juhl Jensen3 and Tudor Oprea1 1 University of New Mexico, Albuquerque, USA; 2 NextMove Scientific Software, Cambridge, UK; 3 Novo Nordisk Foundation Center for Protein Research, Copenhagen, DK We mined ClinicalTrials.gov using the CTTI-AACT db from Duke University for drugs/chemicals and diseases/conditions. Named entitiy recognition (NER) requires specialized tools and expertly curated dictionaries for comprehensive and high quality results, hence use of NextMove Leadmine for chemical NER and JensenLab Tagger for disease NER. Study designs and outcomes can offer new and unique drug target knowledge. Target hypotheses can be inferred indirectly via drugs and diseases, a valuable source of evidence to Illuminate the Druggable Genome. Drug trials produce new and rich experimental data AACT accessed Dec 3, 2019. Chemical NER with NextMove LeadMine https://www.nextmovesoftware.com/ Intervention names and study descriptions mined by LeadMine v3.14.1. Drug trials (interventional): 130740; drug names: 14969; SMILES: 4869. Many non-structures, e.g. "placebo", "test product", "medication", "chemotherapy". Top drugs by total mentions: Compounds mapped to: PubChem via PUG REST API, SMILES exact search; ChEMBL via REST API, InChIKey search; targets via ChEMBL bioassays. Targets mapped to IDG-TCRD/Pharos via UniProt ID. Compound-target mapping via PubChem & ChEMBL Overview IDG F2F Meeting - 11-12 Feb 2020 - Arlington, VA This work supported by US National Institutes of Health (grants U54 CA189205 and U24 224370), Illuminating the Druggable Genome Knowledge Management Center (IDG KMC), and by the Novo Nordisk Foundation (grant number NNF14CC0001). Disease NER with JensenLab Tagger https://github.com/larsjuhljensen/tagger JensenLab diseases dictionary, on detailed descriptions. Disease mention totals by merging to resolved Disease Ontology term (DOID). Top diseases by total mentions: doid N_mentions terms DOID:162 31143 CANCER; CANcer; Cancer; Malignant Tumor; Malignant neoplasm; Malignant tumor; Primary Cancer; Primary cancer; cancer; ... DOID:9351 18955 DIABETES; DIABETES MELLITUS; DIAbetes; DIabetes; Diabetes; Diabetes Mellitus; Diabetes mellitus; diabetes; diabetes Mellitus... DOID:6713 18461 CVA; Cerebrovascular Accident; Cerebrovascular Disease; Cerebrovascular accident; Cerebrovascular disease; STROKE; STRokE;… DOID:2030 13621 ANXIETY; Anxiety; Anxiety Disorder; Anxiety State; Anxiety disorder; Anxiety state; anxiety; anxiety disorder; anxiety state; ... DOID:1612 11586 BREAST CANCER; BReast CAncer; BReast Cancer; Breast Cancer; Breast cancer; Breast tumor; Breast-cancer; Primary breast cancer;… DOID:2841 10773 ASTHMA; Asthma; BHR; Bronchial hyper-reactivity; Bronchial hyperreactivity; EIA; Exercise-induced asthma; asthma; bronchial hyper re … DOID:3083 10726 CHRONIC OBSTRUCTIVE PULMONARY DISEASE; COLD; COPD; COPd; Chronic Obstructive Lung Disease; Chronic Obstructive L… DOID:9970 10193 OBESITY; OBesity; Obesity; obEsity; obe-sity; obesity DOID:10763 9816 HBP; HTN; HYPERTENSION; High Blood Pressure; High blood pressure; High-blood pressure; Hypertension; Hypertensive disease; high… Aggregate Analysis of ClinicalTrials.gov (AACT) db from the Clinical Trials Transformation Initiative (CTTI) Why not target NER? Clinical trials not designed to communicate molecular mechanisms to research scientists, but with focus on clinical efficacy and safety. In due diligence we performed target NER, and compared with target NER on arbitrary non-biomedical text: tweets from the Twitter API for #brexit (26 Nov 2019). We find 8.64 target entities per 1000 chars in the tweets (e.g. "TAX", "LIAR", "NHS", "DANGER", "insulin"), vs. 6.63 in the clinical trials descriptions. While not proof this does support prior belief and target inference via chemical NER. nct_id drug_name cid disease_term doid gene_symbol uniprot idgTDL NCT02600741 Fluphenazine 3372 schizophrenia DOID:5419 MC5R P33032 Tchem NCT00008190 melphalan 460612 acute leukemia DOID:12603 MMP2 P08253 Tchem NCT00003012 methotrexate 126941 breast cancer DOID:1612 KDM4E B2RXH2 Tchem NCT01445522 ABT-888 11960529 lymphoma DOID:0060058 PARP12 Q9H0J9 Tdark NCT03201250 Cabozantinib 25102847 multiple myeloma DOID:9538 ANKK1 Q8NFD2 Tbio https://aact.ctti-clinicaltrials.org/ CDK_smi2img N_mentions names 2787 Abraxane; PACLITAXEL; Paclitaxel; Taxol; abraxane; paclitaxel; taxol 2654 CYCLOPHOSPHAMIDE; Ciclophosphamide; Cyclophosphamid; Cyclophosphamide; ciclophosphamide; cyclophosphamide 2552 CISPLATIN; Cis Platinum; Cis-platinum; Cisplatin; Cisplatine; Cisplatinum; cis Platinum; cis-platinum; cisplatin; cisplatine; cisplatinum