High Resolution CGH analysis is developed1, 2, 3 for improving detection of genomic gains and losses by increasing specificity and sensitivity and at the same time include analysis of chromosome regions (e.g. telomeres) which normally are excluded from CGH analysis, and it includes:
§ Automatic exclusion of certain parts of the chromosomes during fluorochrome level normalization by 1) excluding fluorochrome intensities from known problematic parts of the chromosomes (i.e. telomeres, centromeres and boundary pixels), and 2) for human chromosomes by excluding chromosomes with fluorochrome intensities known to deviate from the normal levels of all other normal chromosomes.
§ Automatic exclusion of unreliable fluorescent intensities. This mostly happens for low intensity pixels at boundaries, telomeres, centromeres, and heterochromatic regions, but unreliable extreme high intensity pixels are excluded as well.
§ Automatic correction for unsuppressed repetitive sequences.
CGH ratio profiles from normal DNA have shown to vary systematically on each side of
ratio 1.0 along the CGH profiles. The systematic variation, which from slide to
slide represents different degrees of the same general pattern, depends on the
laboratory doing the analysis and on the protocol they use.
The systematic variation and confidence is modelled on basis of statistics of representative slides, and used as reference for whether or not ratio variations for a CGH analysis represent possible gains or losses in the test DNA. We use the name “Dynamic Standard Reference Intervals” for this model. It models the average of the systematic deviations along the CGH profiles and the corresponding confidence intervals. Any degree (including reflection in the ratio 1.0 axis) of systematic variation along the CGH profiles for a test slide can then be modelled by scaling the systematic deviations (ratio distances to ratio 1.0) of the model by a constant factor, which fits the model to the ratios of the test slide at locations representing normal test DNA. I.e. the scaling factor is found dynamically for each analysed slide.
When profiles for test and reference DNA are studied individually it is seen that both sets of profiles display a general pattern (but in two different levels at the characteristic deviations) for all chromosomes from case to case (see fig. 1).
The essence of High Resolution CGH analysis is to compare confidence intervals of the slide mean of CGH profiles to the corresponding Dynamic Standard Reference Intervals. The comparison for a case is done at the same level, i.e. 99.5% confidence intervals are compared to 99.5% Dynamic Standard Reference Intervals etc. Each profile point is compared individually as it has its own confidence interval and its own Dynamic Standard Reference Interval.
Dynamic Standard Reference Intervals have the following properties:
§ They have to be generated on basis of statistics of CGH ratios of normal DNA analysed by the same protocol as the slides they are going to be compared to. Preferably each laboratory should generate Dynamic Standard Reference Intervals for each CGH protocol they use on basis of analysis of at least 10 representative slides of 10 metaphases each. The "systematic variation" of the CGH ratios from ratio 1.0 should not be to tiny or too large compared to the laboratory standard for these “representative” slides, and you should select cases that show a similar level of the “systematic variation". If the used protocol often results in CGH cases where the "systematic variations" look like being reflected in the ratio 1.0 axis then do also generate a separate Dynamic Standard Reference Intervals set for this kind of cases. For X an Y chromosome analysis individual sets has to be generated for analysis of female test DNA versus female reference DNA, and for male DNA versus Male DNA.
§ They look like confidence intervals of CGH profiles for an “average” case based on analysis of 10 metaphases. I.e. they reflect that the basic pattern of CGH ratios in a systematic way often deviates from ratio 1.0 (clearly seen on human chromosomes 1p, 19 and 22), and they are especially wide at profile areas where CGH measurements are known to be unreliable (heterochromatic regions, the centromeres and the telomeres).
The explanation why they look like confidence intervals for an “average” case of 10 analysed cells is that the statistics are adjusted to assume that only a maximum of 20 chromosomes (10 metaphases) have contributed to the computation. Otherwise the confidence intervals can get extremely narrow if the statistics is based on e.g. 100 cells.
§ They can be scaled automatically or manually for a case to fit the degree of systematic deviation from ratio 1.0. The automatic scaling requires that there are only a limited number of gains and losses in the test DNA, otherwise there is not much to base the scaling on.
The Possible gains or losses in the test DNA on basis of the length normalised CGH profiles are profile locations where the confidence interval of the slide mean do not overlap the dynamic standard reference interval.
The following considerations should be made when confidence levels are chosen:
§ By choosing narrow confidence intervals (95%) non-aberrant fluctuations (false positives) may be marked as aberrant.
§ By choosing the widest confidence interval (99.9%) small abnormalities may be missed (false negatives). Deletions down to 3 Mbp (in 82% of the cells) have been detected without detection of false positives by using a 99.5% confidence level2).
Pseudo ideograms based on the inverted DAPI counter stain band profiles can be used to define the location of possible gains and losses in the test DNA.
Illustration of high resolution CGH on a CytoVision™ Genetix - Applied Imaging cytogenetic workstation:
A case, with a deletion at the telomere of 2q; all three figures are accompanied by a pseudo ideogram based on the mean inverted counter stain band profiles (DAPI) of the analysed chromosomes.
Left: The two CGH ratio profiles of chromosome 2 of a metaphase. Automatically excluded CGH ratios are marked with grey.
Middle: 99.5% confidence intervals (yellow) versus fixed ratio thresholds 0.8 - 1.20 (black).
Right: 99.5% confidence intervals (yellow) versus the corresponding Dynamic Standard Reference Intervals. The two sets of intervals clearly deviate at the deletion.
Notice how well the two sets of intervals otherwise fits. The Dynamic Standard Reference Intervals have been scaled automatically to fit all chromosome classes, but these characteristic patterns could easily and with great certainty have been fitted manually as well.
The dynamic range of Standard Reference Intervals is further illustrated by this Microsoft PowerPoint presentation or by this pdf-version of the Power Point presentation.
Although the frequency of false positive measurements are profoundly reduced compared to
conventional CGH, aberrations detected at 9p11, 16p11 and 1q21 may sometimes
represent artefacts. They may be normal variations as such
variations have been described in these regions, but recently we have found a
sub-pattern to the characteristic shape of the Dynamic
Standard Reference Intervals.
The sub-pattern correlates to the effect of Cot1-DNA suppression of repetitive sequences, and it results in a CGH-ratio offset for the peri-centromeric regions of 1q, 2p, 7, 9, 10p, 13q, 14q, 15q & 18p, for 11q13, 16p and for the whole chromosomes 17, 19, 22. Most affected are the ratio offsets for 19, 17 and 16p.
The automatic Cot1-DNA correction3 cannot correct for this effect.
This is illustrated by the below 3 sets of Dynamic Standard Reference Intervals or by this Microsoft PowerPoint presentation:
§ Cases having better Cot1-DNA suppression of test-DNA than of reference-DNA get low ratios at the regions mentioned above compared to our routine set of Dynamic Standard Reference Intervals. A set of Dynamic Standard Reference Intervals made by such cases looks like this: Low Std.Ref.Int.
§ Our current routine set of Dynamic Standard Reference Intervals look like this: Normal Std.Ref.Int.
§ Cases having worse Cot1-DNA suppression of reference-DNA than of test-DNA get high ratios at the regions mentioned above compared to our routine set of Dynamic Standard Reference Intervals. A set of Dynamic Standard Reference Intervals made by such cases looks like this: High Std.Ref.Int.
1) Kirchhoff M, Gerdes T, Rose H, Maahr J, Ottesen AM, Lundsteen C. 1998. Detection of chromosomal gains and losses in comparative genomic hybridization analysis based on standard reference intervals. Cytometry 31:163-183.
2) Kirchhoff M, Gerdes T, Maahr J, Rose H, Bentz M, Döhner H, Lundsteen C. 1999. Deletions below 10 Megabasepairs are detected in comparative genomic hybridization by standard reference intervals. Genes Chrom Cancer 25:410-413.
3) Kirchhoff M, Gerdes T, Maahr J, Rose H, Lundsteen C. 1997. Automatic correction of the interfering effect of unsuppressed interspersed repetitive sequences in comparative genomic hybridization analysis. Cytometry 28:130-134.