Skip to main content

Lund University Publications

LUND UNIVERSITY LIBRARIES

Microarray image analysis: background estimation using quantile and morphological filters

Bengtsson, Anders LU and Bengtsson, Henrik LU (2006) In BMC Bioinformatics 7.
Abstract
Background: In a microarray experiment the difference in expression between genes on the same slide is up to 103 fold or more. At low expression, even a small error in the estimate will have great influence on the final test and reference ratios. In addition to the true spot intensity the scanned signal consists of different kinds of noise referred to as background. In order to assess the true spot intensity background must be subtracted. The standard approach to estimate background intensities is to assume they are equal to the intensity levels between spots. In the literature, morphological opening is suggested to be one of the best methods for estimating background this way. Results: This paper examines fundamental properties of rank... (More)
Background: In a microarray experiment the difference in expression between genes on the same slide is up to 103 fold or more. At low expression, even a small error in the estimate will have great influence on the final test and reference ratios. In addition to the true spot intensity the scanned signal consists of different kinds of noise referred to as background. In order to assess the true spot intensity background must be subtracted. The standard approach to estimate background intensities is to assume they are equal to the intensity levels between spots. In the literature, morphological opening is suggested to be one of the best methods for estimating background this way. Results: This paper examines fundamental properties of rank and quantile filters, which include morphological filters at the extremes, with focus on their ability to estimate between-spot intensity levels. The bias and variance of these filter estimates are driven by the number of background pixels used and their distributions. A new rank-filter algorithm is implemented and compared to methods available in Spot by CSIRO and GenePix Pro by Axon Instruments. Spot's morphological opening has a mean bias between -47 and -248 compared to a bias between 2 and -2 for the rank filter and the variability of the morphological opening estimate is 3 times higher than for the rank filter. The mean bias of Spot's second method, morph. close. open, is between -5 and -16 and the variability is approximately the same as for morphological opening. The variability of GenePix Pro's region-based estimate is more than ten times higher than the variability of the rank-filter estimate and with slightly more bias. The large variability is because the size of the background window changes with spot size. To overcome this, a non-adaptive region-based method is implemented. Its bias and variability are comparable to that of the rank filter. Conclusion: The performance of more advanced rank filters is equal to the best region-based methods. However, in order to get unbiased estimates these filters have to be implemented with great care. The performance of morphological opening is in general poor with a substantial spatial-dependent bias. (Less)
Please use this url to cite or link to this publication:
author
and
organization
publishing date
type
Contribution to journal
publication status
published
subject
in
BMC Bioinformatics
volume
7
publisher
BioMed Central (BMC)
external identifiers
  • wos:000239515700001
  • scopus:33746741041
ISSN
1471-2105
DOI
10.1186/1471-2105-7-96
language
English
LU publication?
yes
id
abd85973-5f15-4406-8d9c-bf24dd4fc35a (old id 908528)
date added to LUP
2016-04-01 17:13:46
date last changed
2021-09-22 04:59:05
@article{abd85973-5f15-4406-8d9c-bf24dd4fc35a,
  abstract     = {Background: In a microarray experiment the difference in expression between genes on the same slide is up to 103 fold or more. At low expression, even a small error in the estimate will have great influence on the final test and reference ratios. In addition to the true spot intensity the scanned signal consists of different kinds of noise referred to as background. In order to assess the true spot intensity background must be subtracted. The standard approach to estimate background intensities is to assume they are equal to the intensity levels between spots. In the literature, morphological opening is suggested to be one of the best methods for estimating background this way. Results: This paper examines fundamental properties of rank and quantile filters, which include morphological filters at the extremes, with focus on their ability to estimate between-spot intensity levels. The bias and variance of these filter estimates are driven by the number of background pixels used and their distributions. A new rank-filter algorithm is implemented and compared to methods available in Spot by CSIRO and GenePix Pro by Axon Instruments. Spot's morphological opening has a mean bias between -47 and -248 compared to a bias between 2 and -2 for the rank filter and the variability of the morphological opening estimate is 3 times higher than for the rank filter. The mean bias of Spot's second method, morph. close. open, is between -5 and -16 and the variability is approximately the same as for morphological opening. The variability of GenePix Pro's region-based estimate is more than ten times higher than the variability of the rank-filter estimate and with slightly more bias. The large variability is because the size of the background window changes with spot size. To overcome this, a non-adaptive region-based method is implemented. Its bias and variability are comparable to that of the rank filter. Conclusion: The performance of more advanced rank filters is equal to the best region-based methods. However, in order to get unbiased estimates these filters have to be implemented with great care. The performance of morphological opening is in general poor with a substantial spatial-dependent bias.},
  author       = {Bengtsson, Anders and Bengtsson, Henrik},
  issn         = {1471-2105},
  language     = {eng},
  publisher    = {BioMed Central (BMC)},
  series       = {BMC Bioinformatics},
  title        = {Microarray image analysis: background estimation using quantile and morphological filters},
  url          = {http://dx.doi.org/10.1186/1471-2105-7-96},
  doi          = {10.1186/1471-2105-7-96},
  volume       = {7},
  year         = {2006},
}