Update README.md

README.md

@@ -1,20 +1,47 @@
 ---
-title: Flexynesis Tissue Vae
-emoji: 🚀
-colorFrom: red
-colorTo: red
-sdk: docker
-app_port: 8501
-tags:
-- streamlit
+title: Flexynesis Tissue VAE
+emoji: 🧬
+colorFrom: blue
+colorTo: purple
+sdk: streamlit
+sdk_version: 1.32.0
+app_file: app.py
 pinned: false
-short_description: Flexynesis module for bulk RNA seq
 license: mit
 ---
 
-# Welcome to Streamlit!
+# Flexynesis Tissue VAE
 
-Edit `/src/streamlit_app.py` to customize this app to your heart's desire. :heart:
+Upload a bulk RNA-seq gene expression matrix → get UBERON tissue classification + 121-dimensional latent embeddings.
 
-If you have any questions, checkout our [documentation](https://docs.streamlit.io) and [community
-forums](https://discuss.streamlit.io).
+## Model
+
+Supervised variational autoencoder trained on **75,619 samples** from TCGA, GTEx, DepMap, and ARCHS4 across **43 UBERON tissue categories**.
+
+- **Balanced accuracy:** 90.7%
+- **Weighted F1:** 93.7%
+- **Latent space:** 121 dimensions
+- **Genes:** 16,121 HGNC symbols
+
+## Usage
+
+1. Upload a CSV/TSV with gene expression values (genes × samples or samples × genes)
+2. Gene symbols must be HGNC format
+3. Values should be log2-transformed (TPM, RPKM, or counts)
+4. Click **Classify** to get tissue predictions and embeddings
+
+## Output
+
+- Predicted UBERON tissue category per sample
+- Classification confidence (softmax probability)
+- kNN source breakdown (nearest training samples by data source)
+- Downloadable 121-dim latent embeddings (CSV)
+
+## Citation
+
+Pande A, Uyar B, Franke V, Akalin A. Supervised variational autoencoders learn tissue-specific transcriptomic representations across heterogeneous bulk RNA-seq compendia. *bioRxiv* (2026).
+
+## About
+
+Akalin Lab, Max Delbrück Center for Molecular Medicine (MDC) Berlin/BIMSB  
+GitHub: https://github.com/BIMSBbioinfo/flexynesis