Acquisition of IR spectra often generates complex datasets that are not readily interpretable for the purposes of deriving biomarkers. From a computational perspective, this raises the question of what multi-step processing is required and, whether there is a well-defined sequence of steps that can be applied to
objectively shed insight into a biological question. To generate a dataset to investigate this, we set up an in vitro transformation assay (pH 6.7) using Syrian hamster embryo (SHE) cells (1). SHE cells were interrogated by ATR-FTIR spectroscopy.
Derived mid-IR spectra (nspectra @14,000) were inputted into a computational framework designed for outlier removal, multivariate analysis and validation of the robustness of analysis, and biomarker identification. Biomarker identification
methods were independently applied and compared to identify common discriminating chemical entities. Stable biomarkers of chemical-induced alterations or transformation were identified and confirmed. The analysis framework was implemented in the form of a user-friendly graphical user interface using a programming toolkit designed for research on computational
methods. The database platform developed to store our dataset is scalable and can facilitate a data-sharing inter-laboratory process towards end-user applications for IR spectroscopy.