Fire debris analysis involves interpreting gas chromatography-mass spectrometry (GC-MS) data to determine whether a sample collected from a fire scene contains an ignitable liquid. While all fire debris includes volatile products of substrate materials that decompose during the fire, some samples may also contain an ignitable liquid residue. The volatile chemical compounds in the fire debris are collected and analyzed using GC-MS, which produces a total ion chromatogram (TIC) for evaluation. A TIC provides the relative concentration represented as intensity of the separated chemical compounds throughout the time of analysis. The separated chemical compounds are seen as peaks in the chromatogram at different retention indices, which are related to the time required for each compound to travel the gas chromatograph column to the detector. The TIC data sets contain the intensity of the chemical compounds at each retention index that ranges from 1 to 2800. The total ion spectrum (TIS) datasets contain the average mass spectrum across the chromatographic profile. The rows of a TIC or a TIS dataset represent 60,000 sample observations. The columns of the datasets represent features, which are the intensities of 2800 indices for TIC datasets or the intensities of 131 ions for TIS datasets. In addition, the sample Information dataset files contain the classification designation, labelled class of the fire debris sample, as containing an ignitable liquid (IL) or not containing an ignitable liquid (SUB).
Four datasets of 60,000 insilico fire debris data are available for download. Each dataset contains three files: sample information, total ion chromatograms, and total ion spectra. The files are in zipped (.zip) comma separated value (csv) file formats.
The Fire Debris Database dataset contains sample information, total ion chromatograms, and total ion spectra from the sample records in the Fire Debris Database. These are laboratory generated fire debris samples rather than electronically generated like the insilico fire debris in the other datasets.
More information is available in the article In silico created fire debris data for Machine learning and the training videos below.
Sigman, M.E., Williams, M.R., Tang, L., Booppasiri, S., Prakash, N., In silico created fire debris data for Machine learning, Forensic Chemistry, 42 (2025) 100633, https://doi.org/10.1016/j.forc.2024.100633