HOW_TO_RUN_MMAF.txt

# How to Run MMAF Experiments

## 1. Prepare the project
Put the project folder in Google Colab/Drive, VS Code, or your local machine.

Your project should look like this:

MMAF_Faiz/
  code/
  config/
  data/
  results/
  output/
  requirements.txt

## 2. Add datasets

Put all dataset CSV files inside the data/ folder using lowercase names:

data/tox21.csv
data/bace.csv
data/bbbp.csv
data/hiv.csv
data/clintox.csv
data/sider.csv

If your files are uppercase, rename them:

mv data/BACE.csv data/bace.csv
mv data/BBBP.csv data/bbbp.csv
mv data/HIV.csv data/hiv.csv

## 3. Install requirements

Run from the project root:

pip install -r requirements.txt

## 4. Check setup

Before training, check that the required dataset files exist:

python code/check_setup.py

In Google Colab/Jupyter, use:

!python code/check_setup.py

# Dataset Types and Commands

## A. Tox21

python code/train_tox21.py --config config/tox21.yaml

In Colab:

!python code/train_tox21.py --config config/tox21.yaml

## B. Binary classification datasets

These datasets use the binary pipeline:

- BACE
- BBBP
- HIV

### BACE

Random split:

python code/train_bace.py --split random

Scaffold split:

python code/train_bace.py --split scaffold

Direct pipeline commands:

python code/pipelines/binary_random.py --dataset bace --data-dir data --output-dir results
python code/pipelines/binary_scaffold.py --dataset bace --data-dir data --output-dir results

### BBBP

Random split:

python code/train_bbbp.py --split random

Scaffold split:

python code/train_bbbp.py --split scaffold

Direct pipeline commands:

python code/pipelines/binary_random.py --dataset bbbp --data-dir data --output-dir results
python code/pipelines/binary_scaffold.py --dataset bbbp --data-dir data --output-dir results

### HIV

Random split:

python code/train_hiv.py --split random

Scaffold split:

python code/train_hiv.py --split scaffold

Direct pipeline commands:

python code/pipelines/binary_random.py --dataset hiv --data-dir data --output-dir results
python code/pipelines/binary_scaffold.py --dataset hiv --data-dir data --output-dir results

## C. Multi-output / multi-label datasets

These datasets use the multi pipeline:

- ClinTox
- SIDER

ClinTox has two output labels:

- FDA_APPROVED
- CT_TOX

SIDER has multiple side-effect labels.

### ClinTox

Random split:

python code/train_clintox.py --split random

Scaffold split:

python code/train_clintox.py --split scaffold

Direct pipeline commands:

python code/pipelines/multi_random.py --dataset clintox --data-dir data --output-dir results
python code/pipelines/multi_scaffold.py --dataset clintox --data-dir data --output-dir results

### SIDER

Random split:

python code/train_sider.py --split random

Scaffold split:

python code/train_sider.py --split scaffold

Direct pipeline commands:

python code/pipelines/multi_random.py --dataset sider --data-dir data --output-dir results
python code/pipelines/multi_scaffold.py --dataset sider --data-dir data --output-dir results

# Quick Test

Before running full experiments, test with a quick run:

python code/train.py --dataset bbbp --split random --quick

or:

python code/run_all.py --quick

In Colab:

!python code/run_all.py --quick

# Run All Experiments

To run all datasets and splits:

python code/run_all.py

In Colab:

!python code/run_all.py

# Important Notes

Use lowercase dataset names only:

bace
bbbp
hiv
clintox
sider

Do not use uppercase names such as:

BACE
BBBP
HIV
CLINTOX
SIDER

For example, this is correct:

python code/pipelines/multi_random.py --dataset clintox --data-dir data --output-dir results

This is wrong:

python code/pipelines/multi_random.py --dataset CLINTOX --data-dir data --output-dir results

# Results

Results are saved under:

results/
output/
output/checkpoints/
output/splits/

For example:

results/clintox_random/
results/sider_random/
results/bbbp_random/
results/hiv_scaffold/

To check saved results:

ls results
ls output

In Colab:

!ls results
!ls output