Limits...
Empirical assessment of competitive hybridization and noise in ultra high density canine tiling arrays.

Willet CE, Bunbury-Cruickshank L, van Rooy D, Child G, Shariflou MR, Thomson PC, Wade CM - BMC Bioinformatics (2013)

Bottom Line: Empirical evaluation of competitive hybridization and an estimation of what other non-sequence related features might affect noisy data is currently lacking.Signal variation not attributable to sequence composition resulted from the reduction in competition when large inter-probe spacing was introduced due to long repetitive elements and when a lower density offset was applied.To prevent the introduction of noise from this source, the degree of competition should be regulated by minimizing variation in density across the target region.

View Article: PubMed Central - HTML - PubMed

Affiliation: Faculty of Veterinary Science, The University of Sydney, Sydney, NSW 2006, Australia. cali.willet@sydney.edu.au

ABSTRACT

Background: In addition to probe sequence characteristics, noise in hybridization array data is thought to be influenced by competitive hybridization between probes tiled at high densities. Empirical evaluation of competitive hybridization and an estimation of what other non-sequence related features might affect noisy data is currently lacking.

Results: A high density array was designed to a 1.5 megabase region of the canine genome to explore the potential for probe competition to introduce noise. Multivariate assessment of the influence of probe, segment and design characteristics on hybridization intensity demonstrate that whilst increased density significantly depresses fluorescence intensities, this effect is largely consistent when an ultra high density offset is applied. Signal variation not attributable to sequence composition resulted from the reduction in competition when large inter-probe spacing was introduced due to long repetitive elements and when a lower density offset was applied. Tiling of probes immediately adjacent to various classes of repeat elements did not generate noise. Comparison of identical probe sets hybridized with DNA extracted from blood or saliva establishes salivary DNA as a source of noise.

Conclusions: This analysis demonstrates the occurrence of competitive hybridization between oligonucleotide probes in high density tiling arrays. It supports that probe competition does not generate random noise when it is maintained across a region. To prevent the introduction of noise from this source, the degree of competition should be regulated by minimizing variation in density across the target region. This finding can make an important contribution to optimizing coverage whilst minimizing sources of noise in the design of high density tiling arrays.

Show MeSH

Related in: MedlinePlus

Predicted mean loge median signal by loge distance between tiled segments (bp). Mean loge median signal at each level of loge distance when all REML covariates are held constant at the mean and averaged over all factor levels (solid line). Dashed lines indicate the mean loge median signal +/- standard errors of the predictions.
© Copyright Policy - open-access
Related In: Results  -  Collection

License
getmorefigures.php?uid=PMC3733988&req=5

Figure 3: Predicted mean loge median signal by loge distance between tiled segments (bp). Mean loge median signal at each level of loge distance when all REML covariates are held constant at the mean and averaged over all factor levels (solid line). Dashed lines indicate the mean loge median signal +/- standard errors of the predictions.

Mentions: Position of probe was found to have a significant linear effect on hybridization intensity (P = 0.044, Table 2, Figure 2), with each unit increase in probe position associated with an increase in fluorescence intensity of 0.0002 ± 0.00009 on the loge scale. Increasing distance in bp between tiled segments displayed a significant positive linear effect on hybridization intensity for probes tiled at the edges of segments (P = 0.047, Table 3, Figure 3), with each additional bp separating tiled segments increasing signal from these probes by 0.024 ± 0.012 on the loge scale. Length of tiled segment did not significantly influence hybridization capacity of probes tiled at the edges of segments (Table 3), but did have a significant negative linear effect on intensity when all probes within a segment were considered (P < 0.001, Table 2; Figure 4, REML estimated effect -0.019 ± 0.004 on the loge scale).


Empirical assessment of competitive hybridization and noise in ultra high density canine tiling arrays.

Willet CE, Bunbury-Cruickshank L, van Rooy D, Child G, Shariflou MR, Thomson PC, Wade CM - BMC Bioinformatics (2013)

Predicted mean loge median signal by loge distance between tiled segments (bp). Mean loge median signal at each level of loge distance when all REML covariates are held constant at the mean and averaged over all factor levels (solid line). Dashed lines indicate the mean loge median signal +/- standard errors of the predictions.
© Copyright Policy - open-access
Related In: Results  -  Collection

License
Show All Figures
getmorefigures.php?uid=PMC3733988&req=5

Figure 3: Predicted mean loge median signal by loge distance between tiled segments (bp). Mean loge median signal at each level of loge distance when all REML covariates are held constant at the mean and averaged over all factor levels (solid line). Dashed lines indicate the mean loge median signal +/- standard errors of the predictions.
Mentions: Position of probe was found to have a significant linear effect on hybridization intensity (P = 0.044, Table 2, Figure 2), with each unit increase in probe position associated with an increase in fluorescence intensity of 0.0002 ± 0.00009 on the loge scale. Increasing distance in bp between tiled segments displayed a significant positive linear effect on hybridization intensity for probes tiled at the edges of segments (P = 0.047, Table 3, Figure 3), with each additional bp separating tiled segments increasing signal from these probes by 0.024 ± 0.012 on the loge scale. Length of tiled segment did not significantly influence hybridization capacity of probes tiled at the edges of segments (Table 3), but did have a significant negative linear effect on intensity when all probes within a segment were considered (P < 0.001, Table 2; Figure 4, REML estimated effect -0.019 ± 0.004 on the loge scale).

Bottom Line: Empirical evaluation of competitive hybridization and an estimation of what other non-sequence related features might affect noisy data is currently lacking.Signal variation not attributable to sequence composition resulted from the reduction in competition when large inter-probe spacing was introduced due to long repetitive elements and when a lower density offset was applied.To prevent the introduction of noise from this source, the degree of competition should be regulated by minimizing variation in density across the target region.

View Article: PubMed Central - HTML - PubMed

Affiliation: Faculty of Veterinary Science, The University of Sydney, Sydney, NSW 2006, Australia. cali.willet@sydney.edu.au

ABSTRACT

Background: In addition to probe sequence characteristics, noise in hybridization array data is thought to be influenced by competitive hybridization between probes tiled at high densities. Empirical evaluation of competitive hybridization and an estimation of what other non-sequence related features might affect noisy data is currently lacking.

Results: A high density array was designed to a 1.5 megabase region of the canine genome to explore the potential for probe competition to introduce noise. Multivariate assessment of the influence of probe, segment and design characteristics on hybridization intensity demonstrate that whilst increased density significantly depresses fluorescence intensities, this effect is largely consistent when an ultra high density offset is applied. Signal variation not attributable to sequence composition resulted from the reduction in competition when large inter-probe spacing was introduced due to long repetitive elements and when a lower density offset was applied. Tiling of probes immediately adjacent to various classes of repeat elements did not generate noise. Comparison of identical probe sets hybridized with DNA extracted from blood or saliva establishes salivary DNA as a source of noise.

Conclusions: This analysis demonstrates the occurrence of competitive hybridization between oligonucleotide probes in high density tiling arrays. It supports that probe competition does not generate random noise when it is maintained across a region. To prevent the introduction of noise from this source, the degree of competition should be regulated by minimizing variation in density across the target region. This finding can make an important contribution to optimizing coverage whilst minimizing sources of noise in the design of high density tiling arrays.

Show MeSH
Related in: MedlinePlus