MIALAB

scYOU Server

About scYOU

scYOU is a multi-scale yoked optimization framework that jointly models cellular heterogeneity and protein-level functional structure in proteomics. By integrating supercell-guided alignment with GO-based functional regularization, scYOU couples cell-level organization with protein functional similarity, enabling the learning of coherent and biologically meaningful representations under sparse and noisy measurements.

Submitting Jobs

  • Three data files are required: Proteomics Expression Matrix, GO Similarity Matrix, Supercell Labels.
  • Fourth file (optional): Cell Type Labels for performance evaluation.
  • Adjust hyperparameters: number of clusters, top variable proteins, alpha, beta, gamma, delta, learning rate.

Job Output

You will get a zip file containing:

  • cell_embeddings.csv – Learned cell embeddings.
  • protein_embeddings.csv – Learned protein embeddings.
  • final_cluster_labels.csv – Final cluster assignments for each cell.
  • clustering_results.txt – Evaluation metrics (NMI, ARI) if labels provided.

Download Example

The example contains sample data files to help you understand the correct format.

expression_Montalvo.csv - Protein expression matrix (proteins × cells)
GO_Montalvo.csv - GO similarity matrix (proteins × proteins)
supercell_Montalvo.csv - Supercell labels (one per cell)
meta_Montalvo.csv - Ground truth cell type labels (optional)

Expected output: Cell embeddings, protein embeddings, and cluster labels.
Download Example Files

Upload Data Files

Submit your proteomics datasets for scYOU clustering analysis

Files uploaded successfully!
Please select all required files before submitting.
Protein expression data (proteins × cells) including protein name and cell index
GO similarity matrix (proteins × proteins) including protein name. The GO similarity matrix was constructed using the method described in: Yu G, Li F, Qin Y, Bo X, Wu Y and Wang S. GOSemSim: an R package for measuring semantic similarity among GO terms and gene products. Bioinformatics. 2010, 26(7):976-978.
Supercell labels (single column, column name: supercell_label). The generation of supercells can refer to the scHG method.
Cell labels (single column, column name: label); If provided, performance metrics (NMI, ARI) will be calculated
Weight of reconstruction loss; Default: 1.0
Weight of alignment loss; Default: 0.1
Weight of clustering loss; Default: 1.0
Weight of GO regularization loss; Default: 1.0
Default: 0.001
Reset Form

Upload Status

No file uploaded yet.
Please upload data and set parameters, then click "Run".

Run scYOU Clustering

Execute clustering analysis on uploaded files

Analysis Results

No file uploaded yet.
Please upload data and set parameters, then click "Run".