diff --git a/README.md b/README.md
index 1cacfb6..1478497 100644
--- a/README.md
+++ b/README.md
@@ -1,2 +1,16 @@
 # Molecule-RNN
-Generate molecules for drug discovery.
+Molecule-RNN is a recurrent neural network built with Pytorch to generate molecules for drug discovery. It is trained with the [Zinc](https://zinc.docking.org/) dataset.
+
+## Training
+1. Dowdload the SMILES files of molecules from [Zinc](https://zinc.docking.org/). Select "SMILES(*.smi)" and "Flat" opitions when downloading.
+
+2. Modify the path of dataset in ```train.yaml``` to your downloaded dataset by setting the value of ```dataset_dir```.
+
+3. Run the training script.
+```python train.py```
+
+The training loss:
+
+## Sampling
+We can generate molecules by sampling the model according to the output distribution.
+```python sample.py```
\ No newline at end of file