Cholesterol-carrying lipoprotein particles in blood have long been recognized as key contributors to the development of vascular disease and subsequent incidence of heart attack or stroke. Due to technical difficulties, the information available in clinical practice has been limited to total cholesterol concentration, cholesterol content of high-density lipoproteins (HDL-C) and total concentration of triglycerides. For this reason, a mathematical formula (by Friedewald et al.) for the biologically interesting low-density lipoprotein (LDL-C) was derived from the three basic quantities already in the seventies.
Strictly speaking, the Friedewald formula does not estimate pure LDL-C concentration, but the sum of intermediate density lipoproteins (IDL-C) and LDL-C, with small amount of Lp(a). Lipido is a software that aims to extend the rationale of Friedewald to other important lipid and lipoprotein markers such as triglycerides in very low-density liporoteins (VLDL-TG) and apolipoproteins. The software also calculates estimates for pure LDL-C and IDL-C, thus improving the specificity of the Friedewald formula.
This work spawned from a collaborative project between a number of Finnish study groups. The Computational Medicine Research Group (formely within Helsinki University of Technology, now affiliated with University of Oulu) created the neural network models from an extensive training set of lipid measurements. The results were then validated in the FinnDiane set of type 1 diabetic patients. The computational models involved are very complicated; we have therefore created an easy-to-use online interface, which should provide adequate service to all interested. Our resources are limited, but we are nevertheless interested in any feedback that might lead to better applications.
Computational Medicine Research Group
University of Oulu and Biocenter Oulu
Department of Biomedical Engineering and Computational Science
Aalto University, School of Science and Technology
Finnish Diabetic Nephropathy Study
Folkhälsan Institute of Genetics
Estimation of VLDL, IDL, LDL, HDL2, ApoA-I and ApoB from the Friedewald inputs – ApoB and IDL, but Not LDL, Are Associated with Mortality in Type 1 Diabetes, in press
Current code is compatible with GNU Octave 3.0.1 on Ubuntu Linux.
2008-03-07 Lipido online
The estimation models are defined as Matlab-compatible data structures for proprietary neural network software. Therefore, no source code is available for public distribution.