An In Silico Approach to Finding the Expected Frequency of Coincidental Overlaps for In Situ Hybridization Using Dual-Colour Fusion Probes
Michael Bonert, Kenneth J Craddock. University of Toronto, Toronto, ON, Canada; University Health Network, Toronto, ON, Canada
Background: In situ hybridization (ISH) is a commonly used technique, frequently used to assess for the presence of chromosomal translocations using two-colour fusion-type molecular probes. The success of the technique depends on the detection of (abnormal) signals not seen in normal cells. The coincidental overlapping of probes may mimic an abnormal fusion pattern, and it is generally distinguished from a true pathologic finding by employing statistical methods based on the frequency of these events in normal controls. However, the difference in nuclear size between normal cells and tumour cells may cause the cutoffs to be inaccurate for borderline cases.
Design: The frequency of overlap patterns was calculated using a spherical in silico model of the cell nucleus and random probe locations, using GNU Octave. The parameters in the model include (1) nuclear size (Dn), (2) the maximal distance between the centres of two signals in order to call an overlap in the image plane (res), and (3) orthogonal to the image plane (z_res), and (4) the number of nuclei to randomly generate. Factors (1), (2) and (3) were varied to understand their affect on frequency of overlaps and compared to experimental frequency data of normal controls scored manually at the microscope. A random distribution of the probes was assumed.
Results: The in silico model reproduces the frequency trends of overlaps in the experimental data, and furthermore, allows extrapolation to any given nuclear size. Larger nuclear size is associated with a decrease in overlaps The frequencies of common overlap patterns are functions of a non-dimensional number (res^2*z_res/Dn^3).
Conclusions: The in silico model demonstrates that in a tumour with large nuclei, the expected frequency of various overlap signal patterns is overestimated, if based on counts from smaller (normal) nuclei. This model can be used to determine the expected number of cells with coincidental overlaps for a tumour with any given average nuclear size. Employing criteria derived from this approach should increase the sensitivity of ISH tests uing dual-colour fusion probes.
Monday, March 19, 2012 9:00 AM
Platform Session: Section H, Monday Morning