finish readme

Repeat/CoMPT/README.md

@@ -77,7 +77,7 @@ This result could be refer from [GCN](https://jcheminf.biomedcentral.com/article
 
 *(3) I guess the worse performance in the reimplementation may be the preprocess node and the edge features, the second reason may be due to the remove of some residual connection.*
 
-*(4) The chemical shift prediction may be no need of edge hidden update, you can refer the line 271 for model_graph.py and model_node.py, otherwise will affect the final performance.*
+*(4) The chemical shift prediction may be no need of edge hidden update, you can refer the line 271 for `model_graph.py` and `model_node.py`, otherwise will affect the final performance.*
 
 *(5) I have tried to add the gru layer in the pooling module and found the performance of chemical shift prediction drop a lot, while this layer will increase the protein binding site prediction(A CoMPT version optimized by protein), so there will be more works on it.*