Data Science: Similarity Least Squares (SLSTM) + physics + Statistical Design of Experiment (DOE)

            A New Paradigm for Analysis and Management of Complexity

horizontal rule

HCV prognostics
Clinical study

Viral load drop is a sensitive indicator of sustained response to therapy. Our analysis shows that progressive lowering of viral load over the course of treatment is a powerful prognostic. 

In addition to viral load, other clinical factors impact clinical outcome. These other factors become especially important when a patient's viral-load profile is non-optimal for sustained response prognostics. The SLSTM model leverages these other factors effectively and efficiently with the help of DOE concepts.

In general new clinical diagnostics are becoming increasingly sensitive for final clinical outcome, but specificity tends to remain mediocre.  The SLSTM system produces prognostic models that not only promotes this inherent sensitivity but also advances specificity to acceptable levels by utilizing all the information content in the patient's clinical profile.
For the first time one trains (optimizes/fits) clinical outcome on 30 patients and then can predict accurately
the outcome on 150 other patients. These "critical" patients define the boundary of successful clinical outcome.

The application/impact of SLSTM to other fields such as toxicology would be evolutionary.
The SLSTM process applied to available clinical HCV data:

  1. Build model including functions that cover important combinations among clinical variables, ie, the super vectors.
  2. Identify critical patients, a DOE-based process
  3. Produce a stable model trained on the critical patients
  4. Validate the model

horizontal rule

Copyright of James M Minor, July 4, 2004.
For problems or questions regarding this web contact email
Last updated: June 20, 2013.