Chemometric Analysis

What is Chemometric Analysis?

Molecule Icon

Chemometrics is the application of advanced multivariate statistical methods to chemical analyses. Principal Components Analysis (PCA) is a powerful statistical method for dealing with the massive amounts of data generated by modern instrumentation. SIMCA, a subset of PCA, is particularly useful in reducing many variables to a few key variables or combination of variables (latent variables) which will discriminate between two or more classes of compounds.

Research at MUFSC

test tubes and beakers in laboratory

Currently research is supported by a NIJ CLIP grant applying PCA to the analysis of smokeless powders utilized in Improvised Explosive Devices (IEDs or “pipe bombs”). This work is in cooperation with members of the Technical Working Group for Fire and Explosives (TWGFEX), which is establishing a database of analyses on smokeless powders.

The goal of the research is the individualization of the smokeless powder down to the lot level, allowing law enforcement personnel to better generate leads in crimes involving IEDs. Three forms of analyses (GCMS, HPLC, and CE) will be combined with measurements of grain size and shape to create a data library which can be studied via PCA. Additionally this database can be used to establish a probability of random matches similar to statistical probability utilized in DNA analyses. Future research will include post-blast residue of the same collection of smokeless powders.

Working with data from the ATF

Another project involves individualization of gasoline from fire debris. PCA has been applied to a dataset of GCMS analyses obtained in cooperation with forensic chemists at the Bureau of Alcohol, Tobacco and Firearms (ATF). A collection of over 150 gasoline samples has been assembled in the lab from across the country, with initial GCMS analyses. A larger collection of 2000 gasoline samples representing various brands, grades and seasonal blends is being established. Like the smokeless powder research, a large database of analyses is needed to provide a statistical probability to a match between two or more samples. A key part of this project will be to validate these methods for simulated and actual fire debris.