The scientific literature contains an ever-growing number of reports of applications of vibrational spectroscopy as a multivariate non-invasive tool for analysis of biological effects at the molecular level. Recently, Fourier transform infrared microspectroscopy (FTIRM) has been demonstrated to be sensitive to molecular events occurring in cells and tissue after exposure to ionizing radiation. In this work the application of FTIRM in the examination of dose-dependent molecular effects occurring in skin cells after exposure to ionizing radiation with the use of partial least-squares regression (PLSR) and generalized regression neural networks (GRNN) was studied. The methodology is shown to be sensitive to molecular events occurring with radiation dose and time after exposure. The variation in molecular species with dose and time after irradiation is shown to be non-linear by virtue of the higher modeling efficiency yielded from the non-linear algorithms. Dose prediction efficiencies of approximately +/-10 mGy were achieved at 96 h after irradiation, highlighting the potential applications of the methodology in radiobiological dosimetry.