The KOMAP pipeline comprises two key components:

1. Online Narrative and Codified feature Search engine (ONCE), powered by multi-source knowledge graph representation, and illustrated in the ONCE app;
2. Online phenotyping algorithm training and validation:
   • Training:
     • KOMAP can train a multimodal phenotyping algorithm fully online based on a user-supplied summary of the feature matrix;
   • Validation:
     • KOMAP contains an online evaluation system to approximate evaluation metrics based on additional summary statistics derived from a validation set consisting of labeled data.

• Training:

  With a given set of selected features, including a set of main surrogate features which can be indicated by ONCE and a healthcare utilization measure, the training of KOMAP contains three steps:

  1. Normalizing the main surrogates with the utilization;
  2. Denoising via regression on each main surrogate;
  3. Combining the derived risk scores of different surrogates.
  All of steps only require the embedding of the empirical covariance matrix, FREE of any individual-level information.
• Validation:

  • The key working assumption behind the proposed evaluation algorithm is that all the features given the label follows approximately a Gaussian distribution. With this, the ROC curve of the predicted score is uniquely determined by the conditional mean vectors and conditional covariance matrices.

• Step 0 (Optional): Identify a list of features related to your diease of interests (ONCE);
• Step 1: Upload the training and validation covariance matrix of features with corrupted main surrogates; upload your dictionary connecting variable names to their descriptions;
• Step 2: Specify names for main surrogate(s) and the healthcare utility corresponding to the disease.
• Step 3 (Optional, only needed for simulated AUC): With label data, upload prevalence, conditional mean vectors and conditional covariance matrices;
• Step 4: Just click the "GO KOMAP" button and you are ready to go!