Sublethal effect prediction models

About

This repository contains the code and resources of the following article: Multimodal machine learning for predicting eight sublethal effects resulting from chemical-induced perturbation of 108 physiological/biochemical indicators across multiple fish species

Overview of the framework

We developed a multimodal learning model that integrates chemical structure, species, toxicokinetics, bioactivity, and environmental features to predict the no observed effect concentration (NOEC) values ​​of eight sublethal effects (biochemistry, development, genetics, growth, histology, hormones, morphology, and reproduction) resulting from chemical-induced perturbation of 108 physiological/biochemical indicators across multiple fish species.

Please refer to the following steps to run the sublethal effect models

Running biochemical effect model:

python main.py --effect biochemistry --data data/Biochemistry.csv --target-col logy

Running development effect model:

python main.py --effect development --data data/Development.csv --target-col logy

Running genetics effect model:

python main.py --effect genetics --data data/Genetics.csv --target-col logy

Running growth effect model:

python main.py --effect growth --data data/Growth.csv --target-col logy

Running histology effect model:

python main.py --effect histology --data data/Histology.csv --target-col logy

Running hormones effect model:

python main.py --effect hormones --data data/Hormones.csv --target-col logy

Running morphology effect model:

python main.py --effect morphology --data data/Morphology.csv --target-col logy

Running reproduction effect model:

python main.py --effect reproduction --data data/Reproduction.csv --target-col logy