metadata

license: mit
language:
  - en
tags:
  - chemistry
  - medicinal chemistry
pretty_name: AggregatorAdvisor
size_categories:
  - 10K<n<100K
dataset_summary: >-
  AggregatorAdvisor identifies molecules that are known to aggregate or may
  aggregate in biochemical assays based on the chemical similarity to known
  aggregators, and physical properties. In the default affinity range of 100 nM
  to 10 μM, if calculated LogP > 3 and Tc ≥ 85%, the user is informed that this
  compound should be investigated as an aggregator.  If either calculated LogP >
  3 or Tc > 85%, the user is warned that one of these two contributing criteria
  are in effect and that controls should be run.  If neither of these is true,
  this is reported, and the user is counseled that controls are always advised.
  The train and test datasets were created after sanitizing and splitting the
  original dataset in the paper below.
citation: |-
  @article 
    {Irwin2015, title = {An Aggregation Advisor for Ligand Discovery},
    volume = {58}, ISSN = {1520-4804}, 
    url = {http://dx.doi.org/10.1021/acs.jmedchem.5b01105}, 
    DOI = {10.1021/acs.jmedchem.5b01105}, 
    number = {17}, 
    journal = {Journal of Medicinal Chemistry}, 
    publisher = {American Chemical Society (ACS)}, 
    author = {Irwin, John J. and Duan,  Da and Torosyan,  Hayarpi and Doak,  Allison K. and
    Ziebart,  Kristin T. and Sterling,  Teague and Tumanian,  Gurgen and Shoichet,  Brian K.}, 
    year = {2015}, 
    month = aug, 
    pages = {7076–7087} 
    }
config_names:
  - AggregatorAdvisor
configs:
  - config_name: AggregatorAdvisor
    data_files:
      - split: test
        path: AggregatorAdvisor/test.csv
      - split: train
        path: AggregatorAdvisor/train.csv
dataset_info:
  - config_name: AggregatorAdvisor
    features:
      - name: new SMILES
        dtype: string
      - name: substance_id
        dtype: string
      - name: aggref_index
        dtype: int64
      - name: logP
        dtype: float64
      - name: reference
        dtype: string
    splits:
      - name: train
        num_bytes: 404768
        num_examples: 10116
      - name: test
        num_bytes: 101288
        num_examples: 2529

Aggregator Advisor

The train and test datasets uploaded to our Hugging Face repository have been sanitized and split from the original dataset, which contains 12645 compounds. If you want to try these processes with the original dataset, please follow the instructions in the Processing Script.py file located in the AggregatorAdvisor. The raw_data.csv is the original dataset from the paper, and the files in AggregatorAdvisor are the sanitized version files that we made.

Quickstart Usage

Load a dataset in python

Each subset can be loaded into python using the Huggingface datasets library. First, from the command line install the datasets library

$ pip install datasets

then, from within python load the datasets library

>>> import datasets

and load one of the AggregatorAdvisor datasets, e.g.,

>>> AggregatorAdvisor = datasets.load_dataset("maomlab/AggregatorAdvisor", name = "AggregatorAdvisor")
Downloading readme: 100%|██████████| 4.70k/4.70k [00:00<00:00, 277kB/s]
Downloading data: 100%|██████████| 530k/530k [00:00<00:00, 303kB/s]
Downloading data: 100%|██████████| 2.16M/2.16M [00:00<00:00, 12.1MB/s]
Generating test split: 100%|██████████| 2529/2529 [00:00<00:00, 29924.07 examples/s]
Generating train split: 100%|██████████| 10116/10116 [00:00<00:00, 95081.99 examples/s]

and inspecting the loaded dataset

>>> AggregatorAdvisor
DatasetDict({
test: Dataset({
    features: ['new SMILES', 'substance_id', 'aggref_index', 'logP', 'reference'],
    num_rows: 2529
})
train: Dataset({
    features: ['new SMILES', 'substance_id', 'aggref_index', 'logP', 'reference'],
    num_rows: 10116
})

})

Use a dataset to train a model

One way to use the dataset is through the MolFlux package developed by Exscientia. First, from the command line, install MolFlux library with catboost and rdkit support

pip install 'molflux[catboost,rdkit]'

then load, featurize, split, fit, and evaluate the catboost model

import json
from datasets import load_dataset
from molflux.datasets import featurise_dataset
from molflux.features import load_from_dicts as load_representations_from_dicts
from molflux.splits import load_from_dict as load_split_from_dict
from molflux.modelzoo import load_from_dict as load_model_from_dict
from molflux.metrics import load_suite

Split and evaluate the catboost model

split_dataset = load_dataset('maomlab/AggregatorAdvisor', name = 'AggregatorAdvisor')

split_featurised_dataset = featurise_dataset(
  split_dataset,
  column = "new SMILES",
  representations = load_representations_from_dicts([{"name": "morgan"}, {"name": "maccs_rdkit"}]))

model = load_model_from_dict({
"name": "cat_boost_regressor",
"config": {
    "x_features": ['new SMILES::morgan', 'new SMILES::maccs_rdkit'],
    "y_features": ['logP']}})

model.train(split_featurised_dataset["train"])

preds = model.predict(split_featurised_dataset["test"])

regression_suite = load_suite("regression")

scores = regression_suite.compute(
    references=split_featurised_dataset["test"]['logP'],
    predictions=preds["cat_boost_regressor::logP"])

Data splits

Here we have used the Realistic Split method described in (Martin et al., 2018) to split the AggregatorAdvisor dataset.

Citation

J. Med. Chem. 2015, 58, 17, 7076–7087 Publication Date:August 21, 2015 https://doi.org/10.1021/acs.jmedchem.5b01105