Quantitative structure-activity relationship (QSAR) is a computational modeling method to reveal relationships between structural properties of chemical compounds and biological activities. QSAR modeling is essential for drug discovery, but it has many constraints. Ensemble-based machine learning approaches have been used to overcome constraints and obtain reliable predictions. Ensemble learning builds a set of diversified models and combines them. However, the most prevalent random forest and other ensemble approaches in QSAR prediction limit their model diversity to a single subject.

We propose a comprehensive ensemble method that builds multi-subject diversified models and combines them through second-level meta-learning. In addition, we propose an end-to-end neural network-based individual classifier that can automatically extract sequential features from a simplified molecular-input line-entry system (SMILES). The proposed individual model did not show impressive results as a single model, but it was considered the most important predictor when combined, according to the interpretation of meta-learning. The proposed ensemble method consistently outperformed thirteen individual models on 19 bioassay datasets and demonstrated superiority over other ensemble approaches that are limited to a single subject.

• sckikit-learn
• keras
• tensorflow
• pandas
• joblib
• numpy
• RDKit
• pillow

pip install pandas sklearn tensorflow keras ...

• step1. CREATE CV and RESULTS/PROB DATA folders
  *Create fold numbers for CV. (e.g., CV/1, CV2, ..., CV5)

• step2. Download data to DATA folder
  *For the use of alternative data, please use sequtils.py to covert data type to one of followings: smiles, pubchem, ECFP, MACCS

• step3. run all datasets of qsar_classifiy.py (executing expe_runs.py will do)
  *This will create a learning models of SVM, RF, GBM, NNGB, and NN

• step4. run qsar_weightlearn.py for second-level learning

• step5. run qsar_result.py to get final results

(models.py)

• libraries for model training

(sequtils.py)

• libraries for input handling (representation type can be smiles, pubchem, ECFP, and MACCS)

(expe_runs.py)

• run for all datasets of qsar_classifiy.py

(qsar_classify.py)

• first-level individual learning . 5-fold cross-validation for 2-nd level learning . ./CV/
• Other individual methods on final 25% test set for 5 repetative tries . ./RESULTS/PROB/

(qsar_weightlearn.py)

• second-level for combining . read from CV (5-fold) and make output on ./RESULTS/PROB/. 5 repeat

(qsar_result.py)

• mean of 5 repetative tries
• final results