Add logo

README.md

+# AIIM2023
+<img src="./logo/logo.svg">
+
+## Content in each directory:
+
+
+- **documentation**: Instructions to install the needed libraries.
+- **graphData**: Data of DISNET's graph.
+- **metrics**: Training, testing and RepoDB verification ROC & PRC, also loss evolution for training and validation.
+- **models**: Trained models.
+- **results**: Result files of the RepoDB test and the distribution plots (once these are generated).
+- **Code files:**
+  - deepSnapPred (train & test DeepSnap) 
+  - heterograph_construction (build graph)
+  - testRepoDB (validate model using RepoDB) 
+  - topN (getting top N new predictions of a model)
+  - utilities (plotting utilities)
+  - visualizeDistribution (Checks the distribution of the predictions for a group of diseases, check TFG for more information).
+  - visualizeEmbeddings (visualization of embeddings)
+
+## Summary
+Using Graph Neural Networks (GNNs) to predict new drug repurposing hypothesis. The objective is to predict new edges between 
+drugs and diseases (edge inference) or predict the type of edge (drug) between two diseases (edge classification).
+
+One models is generated using DeepSnap library. The model has an encoder-decoder (GraphSAGE-Dot product) architecture,
+the use of different architectures will be studied in the future. This is the generated models and its status:
+- 🟢 **DeepSnap:** Passed verification tests, proved to be functional and works well.
+
+### Symbol legend:
+- 🟢 : Ready to use.
+- 🟡 : Use is not recommended due to major issues.
+- 🟠 : Not usable