ProTS in Depth
The analysis of MALDI mass spectra of biological samples (for example tissue, serum, bodily fluids) is a promising, but complex, avenue for the discovery of clinically relevant biomarkers and tests. The generation of reproducible spectra and identification of statistically significant differences between groups of spectra are complex problems that involve a variety of difficult technical challenges.
|Estimating and subtracting background that can change from spectrum to spectrum, and also within a single spectrum||A reliable method for background estimation and subtraction|
|Detecting mass spectral peaks in the presence of noise that can vary substantially, both within a spectrum and from spectrum to spectrum||A local estimation of the noise strength|
|Aligning mass spectra to correct for shifts in the m/z axis||A reliable method for spectral alignment and assessment of its accuracy|
|Detecting features in the mass spectra which can distinguish between clinical groups||A robust method of feature detection and calculation of statistical significance of features distinguishing between groups of spectra|
|Constructing a robust classifier based on multiple distinguishing features||An approach based on k-nearest neighbor classifiers, leave-N-out cross-validation methods, and independent validation sets|
Biodesix continues to investigate, evaluate, and implement new strategies to increase the power of the ProTS platform in spectral preprocessing, feature detection, and classifier construction and optimization.