About Genomic Tiling

The minimum number of probes depends on the size of the aberration, the number of probes preferred (e.g., 1, 2, 3 or more) to make an aberration call, and the overall noise of the experiment. A good measure for overall noise of an Agilent aCGH experiment is the QC metric DLRSD (derivative log ratio spread, probe-to-probe log ratio noise), and this strongly correlates with DNA quality.

Regarding the maximum number of probes, putting more probes in a given region does not necessarily result in the ability to make smaller aberration calls. (See also Ylstra B et al. Nucleic Acids Research. 2006; 34(2):445–450) for a discussion on functional resolution of CGH experiments. First of all, if a very high-density design is created, it will surely contain probes that have lower scores. All probes in the Agilent HD-CGH database have a predicted performance score (based on T_m (melting temperature), GC content, a hairpin ΔG, sequence complexity, and metrics to measure homology with the rest of the reference genome). The eArray pair-wise reduction algorithm will pick the best HD probes based on the user-selected average HD probe spacing per interval or the total number of HD probes. In the case of a very high-density design, the pair-wise reduction algorithm will no longer be able to return the best HD probes but will return all or most available probes instead. See the FAQ "What is a CGH probe score?” for more information.

Moreover, very high-density designs might have negative consequences for hybridization. The standard Agilent protocol for DNA labeling yields labeled fragments with a size peak around 200 nt. In the case of a very high-density design, probes on the array will compete for the same fragment for hybridization. Having too many probes in the small region will decrease the signal and might result in noisy data. Agilent does not recommend going lower than 150 to 200 bp spacing.

See the figure below for an example of CGH data from probes at various densities using either the HD search or Genomic Tiling. Probes selected from an HD search with up to 200 bp spacing provided log ratios very close to the expected value of -1.

Overall, probes generated using the ‘Genomic Tiling’ function will perform more poorly than probes found in the Agilent HD-CGH Probe Database. Agilent very strongly recommends using the Agilent HD-CGH Probe Database and not the ‘Genomic Tiling’ option. Only for regions where there are not enough HD probes available in the database should ‘Genomic Tiling’ be considered.

All HD probes in the database (except for probes in regions in which no optimal T_m probes exist) have been T_m matched and have a predicted performance score (based on T_m, GC content, a hairpin ΔG, sequence complexity, and metrics to measure homology with the rest of the reference genome). The eArray pair-wise reduction algorithm will pick the best HD probes based on the user-selected average HD probe spacing per interval or the total number of HD probes.

Additionally, during design the HD probes have passed a Tm filter, are annotated such that a user can choose between different similarity filtering options (non-unique probe filter, perfect match filter, or similarity score filter), and if there are catalog probes present in search results they can be preferentially selected. In contrast, using eArray’s ‘Genomic Tiling’ feature probes are picked at a fixed spacing and there is no chance to Tm balance or optimize probes selected for by performance. Probes created should perform no better than those picked at random. The only options to improve probe performance are probe trimming and skipping of repeat masked regions.

See the figure below for an example of CGH data from Agilent HD-CGH probes compared to ‘Genomic Tiling’ probes. The median log2 ratios of the HD-CGH probes are closer to the expected value of -1 with a smaller spread when compared to the ‘Genomic Tiling’ probes.

• For probes selected using the Agilent HD-CGH Probe Database:
  
  All HD probes in the database have been designed to avoid repeat masked regions, are T_m matched (except for probes in regions in which no optimal T_m probes exist) and have a predicted performance score (based on T_m, GC content, a hairpin ΔG, sequence complexity, and metrics to measure homology with the rest of the reference genome). The eArray pair-wise reduction algorithm will pick the best HD probes based on the user-selected average HD probe spacing per interval or the total number of HD probes. Additionally, HD probes can be filtered using a T_m filter, similarity filter (non-unique probe filter, perfect match filter, or similarity score filter), or catalog probes can be preferentially selected.

• For probes selected using ‘Genomic Tiling’:
  
  Probes in GC-rich, high-T_m, and repeat regions can be very problematic. The only options in ‘Genomic Tiling’ are probe trimming and skipping of repeat masked regions. The new probe set can be further improved by filtering out high-T_m or GC-rich probes. T_m and GC% can be obtained from eArray by using the Score Custom Probes utility under the Probe tab. A probe search using the passing probe IDs can then be used to create a new Probe Group.