ChemProt/DrugProt Testing Ground

Testing ground for task 5 from BioCreative VI

Task 5: Text mining chemical-protein interactions (CHEMPROT)

The aim of the CHEMPROT task of BioCreative VI is to promote the development and evaluation of systems that are able to automatically detect in running text (PubMed abstracts) relations between chemical compounds/drug and genes/proteins. We will therefore release a manually annotated corpus, the CHEMPROT corpus, where domain experts have exhaustively labeled: (a) all chemical and gene mentions, and (b) all binary relationships between them corresponding to a specific set of biologically relevant relation types (CHEMPROT relation classes).

Working Environment

• Ubuntu 20.04
• CUDA 11.2
• PyTorch 1.8.1+cu111

Data Structure

data/
├── chemprot_development
│   ├── chemprot_development_abstracts.tsv
│   ├── chemprot_development_entities.tsv
│   ├── chemprot_development_gold_standard.tsv
│   ├── chemprot_development_relations.tsv
│   └── Readme.pdf
├── chemprot_sample
│   ├── chemprot_sample_abstracts.tsv
│   ├── chemprot_sample_entities.tsv
│   ├── chemprot_sample_gold_standard.tsv
│   ├── chemprot_sample_predictions_eval.txt
│   ├── chemprot_sample_predictions.tsv
│   ├── chemprot_sample_relations.tsv
│   ├── guidelines
│   │   ├── CEM_guidelines.pdf
│   │   ├── CHEMPROT_guidelines_v6.pdf
│   │   └── GPRO_guidelines.pdf
│   └── Readme.pdf
├── chemprot_test_gs
│   ├── chemprot_test_abstracts_gs.tsv
│   ├── chemprot_test_entities_gs.tsv
│   ├── chemprot_test_gold_standard.tsv
│   ├── chemprot_test_relations_gs.tsv
│   └── readme_test_gs.pdf
└── chemprot_training
    ├── chemprot_training_abstracts.tsv
    ├── chemprot_training_entities.tsv
    ├── chemprot_training_gold_standard.tsv
    ├── chemprot_training_relations.tsv
    └── Readme.pdf

5 directories, 25 files

Example

PMID

10471277

Title

Probing the salmeterol binding site on the beta 2-adrenergic receptor using a novel photoaffinity ligand,[(125)I]iodoazidosalmeterol.

Abstract

Salmeterol is a long-acting beta2-adrenergic receptor (beta 2AR) agonist used clinically to treat asthma. In addition to binding at the active agonist site, it has been proposed that salmeterol also binds with very high affinity at a second site, termed the "exosite", and that this exosite contributes to the long duration of action of salmeterol. To determine the position of the phenyl ring of the aralkyloxyalkyl side chain of salmeterol in the beta 2AR binding site, we designed and synthesized the agonist photoaffinity label [(125)I]iodoazidosalmeterol ([125I]IAS). In direct adenylyl cyclase activation, in effects on adenylyl cyclase after pretreatment of intact cells, and in guinea pig tracheal relaxation assays, IAS and the parent drug salmeterol behave essentially the same. Significantly, the photoreactive azide of IAS is positioned on the phenyl ring at the end of the molecule which is thought to be involved in exosite binding. Carrier-free radioiodinated [125I]IAS was used to photolabel epitope-tagged human beta 2AR in membranes prepared from stably transfected HEK 293 cells. Labeling with [(125)I]IAS was blocked by 10 microM (-)-alprenolol and inhibited by addition of GTP gamma S, and [125I]IAS migrated at the same position on an SDS-PAGE gel as the beta 2AR labeled by the antagonist photoaffinity label [125I]iodoazidobenzylpindolol ([125I]IABP). The labeled receptor was purified on a nickel affinity column and cleaved with factor Xa protease at a specific sequence in the large loop between transmembrane segments 5 and 6, yielding two peptides. While the control antagonist photoaffinity label [125I]IABP labeled both the large N-terminal fragment [containing transmembranes (TMs) 1-5] and the smaller C-terminal fragment (containing TMs 6 and 7), essentially all of the [125I]IAS labeling was on the smaller C-terminal peptide containing TMs 6 and 7. This direct biochemical evidence demonstrates that when salmeterol binds to the receptor, its hydrophobic aryloxyalkyl tail is positioned near TM 6 and/or TM 7. A model of IAS binding to the beta 2AR is proposed.

Entity mention annotations

CHEMPROT detailed relation annotations

CHEMPROT task Gold Standard data and predictions

• File: chemprot_sample_gold_standard.tsv

• File: chemprot_sample_predictions.tsv

Knowledge Graph results

Resources

BLUE, the Biomedical Language Understanding Evaluation benchmark

Chemical-protein Interaction Extraction via Gaussian Probability Distribution and External Biomedical Knowledge

Scibert

Model file of scibert_scivocab_uncased:

wget "https://s3-us-west-2.amazonaws.com/ai2-s2-research/scibert/pytorch_models/scibert_scivocab_uncased.tar"