v0.1.6: update README

README.md

@@ -1,6 +1,6 @@
 # COMA: efficient structure-constrained molecular generation using COnstractive and MArgin losses
 
-- Latest update: 05 April 2023
+- Latest update: 07 July 2023
 
 <img src="figs/overview_of_COMA.png" alt="thumbnail" width="600px" />
 
@@ -10,13 +10,13 @@ For a given source molecule, COMA generates a novel molecule with more improved
 
 To achieve property improvement and high structural similarity simultaneously, COMA exploits reinforcement learning and metric learning.
 
-For more detail, please refer to Choi, J., Seo, S. & Park, S. **COMA: efficient structure-constrained molecular generation using contractive and margin losses**. *J Cheminform* 15, 8 (2023). https://doi.org/10.1186/s13321-023-00679-y
+For more detail, please refer to J. Choi, S. Seo, and S. Park. **COMA: efficient structure-constrained molecular generation using contractive and margin losses**. *J Cheminform* 15, 8 (2023). https://doi.org/10.1186/s13321-023-00679-y
 
 
 --------------------------------------------------------------------------------------------
 ## SYSTEM REQUIERMENTS: 
 
-- (If GPU is available) COMA requires GPU memory larger than 6GB.
+- (If GPU is available) COMA may require GPU memory larger than 6GB.
   - Available cudatoolkit versions: 10.2, 11.1, and 11.3
 
 - **COMA is only for Python 3.7**
@@ -46,11 +46,19 @@ cd data
 tar -xzvf drd2.tar.gz
 tar -xzvf qed.tar.gz
 tar -xzvf logp04.tar.gz
-tar -xzvf logp06.tar.gz=
+tar -xzvf logp06.tar.gz
 cd ..
 ```
 
-- A training dataset for COMA is generated by using the jupyter notebooks 'generate_triplet_data_{PROPERTY_NAME}.ipynb' in the data directory
+- After decompressing, an user can find the following files and is ready to run the provided scripts.
+  - rdkit_test.txt
+  - rdkit_train_pairs.txt
+  - rdkit_train_src.txt
+  - rdkit_train_tar.txt
+  - rdkit_train_triplet.txt
+  - rdkit_valid.txt
+
+- The details of how to create a triplet dataset are described in the Algorithm S3 of our paper.
 
 
 --------------------------------------------------------------------------------------------