conv_qsar_fast

QSAR/QSPR using descriptor-free molecular embedding, updated for python 3

In the release python3minimalchange you will find the version of code which runs under python3 and closely resembles the original code (as opposed to the newest version). You can find it under releases or switch to tag python3minimalchange (same button as branch switching).

The newest version has been changed dramatically:

• splitting datasets has been removed, now each fold/subset should be kept in a separate file. As a result:
  • file main/main_cv.py has seen major changes
  • file main/data.py has been simplified a lot. No more splitting, and no more checking which set is used - every information is now kept in config files.
  • as a result config files have changed a lot. Unused fields have been removed, new fields appeared:
    • cv - cross-validation or train-val-test?
    • fold1, ..., foldk - cross validation splits if cv: true
    • train, val - train and valid if cv: false
    • test - the test set, always
  • scripts/split_datasets.py includes the code for splitting Abraham, Bradley and Delany datasets (you will find them already calculated and included in data)
  • TOX21 is not cross-valed anymore, instead the split from the challenge is used
• main/test.py has seen major changes, most importantly you get to choose if rocauc and parity plots should be calculated, the plots and jsons with the score are saved. New section TEST and new fields (calculate_parity, calculate_rocauc) appeared in the config files
• each config file has a full information about the dataset (so that you don't need all those ifs when loading the data), new fields appeared:
  • y_index - label index
  • y_label - label name used for titles of plots
  • delimiter - delimiter used in the data format, used by csv.reader
  • skip_line - some datasets have column names in the first line (set true), others don't (set false)
• in the original code TOX21 was not averaged, now it is. New field (averaging) in config files appeared and can have the following values:
  • mean - mean value is calculated and set as label (this was originally used for Abraham, Bradley and Delaney)
  • max - max value is set as label (new stuff, for TOX)
  • median - median value is set as a label
• support for lr in main.core.build_model has been added. With the current setup it is preferable to use an optimiser with adaptive learning rate (ex. Adam) and set the initial learning rate using lr. In such setup lr_func should not be used as it overwrites the lr parameter.
• changes in mutiple files to make the code more readable, removing files that were not used anymore

Requirements

This code relies on Keras for the machine learning framework, Theano for computations as its back-end, and RDKit for parsing molecules from SMILES strings. Plotting is done in matplotlib. All other required packages should be dependencies of Keras, Theano, or RDKit.

• python 3.7.6
• tensorflow 1.14.0
• keras 2.2.4
• theano 1.0.4
• rdkit 2019.09.3.0

Full copy of environment is listed in directory environment

Issues

Even though result on TOX21 are similar to those in the paper, the model does not learn anything on Abraham, Bradley nor Delaney. Learning on fingerprints (ECFP) does not work.

Basic use

This code implements the tensor-based convolutional embedding strategy described in placeholder for QSAR/QSPR tasks. The model architecture, training schedule, and data source are defined in a configuration file and trained using a cross-validation (CV). The basic architecture is as follows:

• Pre-processing to convert a SMILES string into an attributed graph, then into an attributed adjacency tensor
• Convolutional embedding layer, which takings a molecular tensor and produces a learned feature vector
• Optional dropout layer
• Optional hidden densely-connected neural network layer
• Optional dropout layer
• Optional second hidden densely-connected neural network layer
• Linear dense output layer

Models are built, trained, and tested with the command

python conv_qsar_fast/main/main_cv.py conv_qsar_fast/inputs/<input_file>.cfg

Numerous example input files, corresponding the models described in placeholder are included in inputs. These include models to be trained on full datasets, 5-fold CVs with internal validation and early stopping, 5-fold CVs without internal validation, models initialized with weights from other trained models, and multi-task models predicting on multiple data sets. Note that when using multi-task models, the output_size must be increased and the loss function must be custom to ensure NaN values are filtered out if not all inputs x have the full set of outputs y.

Data sets

There are four available data sets in this version of the code contained in data:

1. Abraham octanol solubility data, from Abraham and Admire's 2014 paper.
2. Delaney aqueous solubility data, from Delaney's 2004 paper.
3. Bradley double plus good melting point data, from Bradley's open science notebook initiative.
4. Tox21 data from the Tox21 Data Challenge 2014, describing toxicity against 12 targets.

Because certain entries could not be unambiguously resolved into chemical structures, or because duplicates in the data sets were found, the effective data sets after processing are exported using scripts/save_data.py as coley_abraham.tdf, coley_delaney.tdf, coley_bradley.tdf, coley_tox21.tdf, and coley_tox21-test.tdf.

Model interpretation

This version of the code contains the general method of non-linear model interpretation of assigning individual atom and bond attributes to their average value in the molecular tensor representation. The extent to which this hurts performance is indicative of how dependent a trained model has become on that atom/bond feature. As long as the configuration file defines a model which loads previously-trained weights, the testing routine is performed by

python conv_qsar_fast/main/test_index_removal.py conv_qsar_fast/inputs/<input_file>.cfg

It is assumed that the trained model used molecular_attributes, as the indices for removal are hard-coded into this script.

Suggestions for modification

Overall architecture

The range of possible architectures (beyond what is enabled with the current configuration file style) can be extended by modifying build_model in main/core.py. See the Keras documentation for ideas.

Data sets

Additional .csv data sets can be incorporated by adding an additional elif statement to main/data.py. As long as one column corresponds to SMILES strings and another to the property target, the existing code can be used with minimal modification.

Atom-level or bond-level attributes

Additional atom- or bond-level attributes can be included by modifying utils/neural_fp.py, specifically the bondAttributes and atomAttributes functions. Because molecules are already stored as RDKit molecule objects, any property calculable in RDKit can easily be added.