Limits...
A flexible Bayesian method for detecting allelic imbalance in RNA-seq data.

León-Novelo LG, McIntyre LM, Fear JM, Graze RM - BMC Genomics (2014)

Bottom Line: The proposed model always has a lower type I error rate compared to the binomial test.Consequently, as variant identification improves, the need for DNA controls will be reduced.Filtering does not significantly improve performance and is not recommended, as information is sacrificed without a measurable gain.

View Article: PubMed Central - PubMed

Affiliation: Department of Biological Sciences, Auburn University, 101 Rouse Life Science Building, 36849 Auburn, AL, USA. rmgraze@auburn.edu.

ABSTRACT

Background: One method of identifying cis regulatory differences is to analyze allele-specific expression (ASE) and identify cases of allelic imbalance (AI). RNA-seq is the most common way to measure ASE and a binomial test is often applied to determine statistical significance of AI. This implicitly assumes that there is no bias in estimation of AI. However, bias has been found to result from multiple factors including: genome ambiguity, reference quality, the mapping algorithm, and biases in the sequencing process. Two alternative approaches have been developed to handle bias: adjusting for bias using a statistical model and filtering regions of the genome suspected of harboring bias. Existing statistical models which account for bias rely on information from DNA controls, which can be cost prohibitive for large intraspecific studies. In contrast, data filtering is inexpensive and straightforward, but necessarily involves sacrificing a portion of the data.

Results: Here we propose a flexible Bayesian model for analysis of AI, which accounts for bias and can be implemented without DNA controls. In lieu of DNA controls, this Poisson-Gamma (PG) model uses an estimate of bias from simulations. The proposed model always has a lower type I error rate compared to the binomial test. Consistent with prior studies, bias dramatically affects the type I error rate. All of the tested models are sensitive to misspecification of bias. The closer the estimate of bias is to the true underlying bias, the lower the type I error rate. Correct estimates of bias result in a level alpha test.

Conclusions: To improve the assessment of AI, some forms of systematic error (e.g., map bias) can be identified using simulation. The resulting estimates of bias can be used to correct for bias in the PG model, without data filtering. Other sources of bias (e.g., unidentified variant calls) can be easily captured by DNA controls, but are missed by common filtering approaches. Consequently, as variant identification improves, the need for DNA controls will be reduced. Filtering does not significantly improve performance and is not recommended, as information is sacrificed without a measurable gain. The PG model developed here performs well when bias is known, or slightly misspecified. The model is flexible and can accommodate differences in experimental design and bias estimation.

Show MeSH

Related in: MedlinePlus

Simulated amount of bias vs Type I error rate of the PG model with different levels of misspecification. The x axis represents the simulated amount of bias x = 2(100)(B-0.5) with B in the interval (0.35,0.65). The line labeled “0” represents the type I error rate TIER when q = (1 + 0%)B, the line labeled “1%” represents the TIER when q = (1 + 1%)B, and similarly for the lines labeled “2%”, “5%” and “10%”. Note that the smaller the simulated amount of bias is the lower B is and, therefore, the difference between B and the specified q, is smaller; hence the lower the TIER. This explains why the TIER increases with the simulated amount of bias. The horizontal line (grey) through 0.05 is shown for reference.
© Copyright Policy - open-access
Related In: Results  -  Collection

License 1 - License 2
getmorefigures.php?uid=PMC4230747&req=5

Fig3: Simulated amount of bias vs Type I error rate of the PG model with different levels of misspecification. The x axis represents the simulated amount of bias x = 2(100)(B-0.5) with B in the interval (0.35,0.65). The line labeled “0” represents the type I error rate TIER when q = (1 + 0%)B, the line labeled “1%” represents the TIER when q = (1 + 1%)B, and similarly for the lines labeled “2%”, “5%” and “10%”. Note that the smaller the simulated amount of bias is the lower B is and, therefore, the difference between B and the specified q, is smaller; hence the lower the TIER. This explains why the TIER increases with the simulated amount of bias. The horizontal line (grey) through 0.05 is shown for reference.

Mentions: To understand how bias affects model performance, we further investigated the behavior of the PG model with q = 1/2 and PG model with q = B, for B = 0.5±10%error. The model performs well when there is bias, while the binomial and q = 1/2 perform poorly when there is bias (Figure 2; Additional file 1: Figure S1). The model with q = B controls the type I error rate (2.6%) even when there is bias in the allele-specific read counts. When bias is accounted for but misspecified, the type I error rate depends on the amount of misspecification (Figure 2; Figure 3). Interestingly, when the amount of bias is large misspecification of small amounts (5%) can result in large type I errors. As expected, this appears as slightly asymmetric with respect to the binomial. This is simply due to 1% of 0.65 being a larger absolute amount of bias than 1% of 0.35 (Figure 3). When bias is large and unaccounted for, the PG model with q = 1/2 and the binomial can have dramatic type I error rates (Figure 2).Figure 3


A flexible Bayesian method for detecting allelic imbalance in RNA-seq data.

León-Novelo LG, McIntyre LM, Fear JM, Graze RM - BMC Genomics (2014)

Simulated amount of bias vs Type I error rate of the PG model with different levels of misspecification. The x axis represents the simulated amount of bias x = 2(100)(B-0.5) with B in the interval (0.35,0.65). The line labeled “0” represents the type I error rate TIER when q = (1 + 0%)B, the line labeled “1%” represents the TIER when q = (1 + 1%)B, and similarly for the lines labeled “2%”, “5%” and “10%”. Note that the smaller the simulated amount of bias is the lower B is and, therefore, the difference between B and the specified q, is smaller; hence the lower the TIER. This explains why the TIER increases with the simulated amount of bias. The horizontal line (grey) through 0.05 is shown for reference.
© Copyright Policy - open-access
Related In: Results  -  Collection

License 1 - License 2
Show All Figures
getmorefigures.php?uid=PMC4230747&req=5

Fig3: Simulated amount of bias vs Type I error rate of the PG model with different levels of misspecification. The x axis represents the simulated amount of bias x = 2(100)(B-0.5) with B in the interval (0.35,0.65). The line labeled “0” represents the type I error rate TIER when q = (1 + 0%)B, the line labeled “1%” represents the TIER when q = (1 + 1%)B, and similarly for the lines labeled “2%”, “5%” and “10%”. Note that the smaller the simulated amount of bias is the lower B is and, therefore, the difference between B and the specified q, is smaller; hence the lower the TIER. This explains why the TIER increases with the simulated amount of bias. The horizontal line (grey) through 0.05 is shown for reference.
Mentions: To understand how bias affects model performance, we further investigated the behavior of the PG model with q = 1/2 and PG model with q = B, for B = 0.5±10%error. The model performs well when there is bias, while the binomial and q = 1/2 perform poorly when there is bias (Figure 2; Additional file 1: Figure S1). The model with q = B controls the type I error rate (2.6%) even when there is bias in the allele-specific read counts. When bias is accounted for but misspecified, the type I error rate depends on the amount of misspecification (Figure 2; Figure 3). Interestingly, when the amount of bias is large misspecification of small amounts (5%) can result in large type I errors. As expected, this appears as slightly asymmetric with respect to the binomial. This is simply due to 1% of 0.65 being a larger absolute amount of bias than 1% of 0.35 (Figure 3). When bias is large and unaccounted for, the PG model with q = 1/2 and the binomial can have dramatic type I error rates (Figure 2).Figure 3

Bottom Line: The proposed model always has a lower type I error rate compared to the binomial test.Consequently, as variant identification improves, the need for DNA controls will be reduced.Filtering does not significantly improve performance and is not recommended, as information is sacrificed without a measurable gain.

View Article: PubMed Central - PubMed

Affiliation: Department of Biological Sciences, Auburn University, 101 Rouse Life Science Building, 36849 Auburn, AL, USA. rmgraze@auburn.edu.

ABSTRACT

Background: One method of identifying cis regulatory differences is to analyze allele-specific expression (ASE) and identify cases of allelic imbalance (AI). RNA-seq is the most common way to measure ASE and a binomial test is often applied to determine statistical significance of AI. This implicitly assumes that there is no bias in estimation of AI. However, bias has been found to result from multiple factors including: genome ambiguity, reference quality, the mapping algorithm, and biases in the sequencing process. Two alternative approaches have been developed to handle bias: adjusting for bias using a statistical model and filtering regions of the genome suspected of harboring bias. Existing statistical models which account for bias rely on information from DNA controls, which can be cost prohibitive for large intraspecific studies. In contrast, data filtering is inexpensive and straightforward, but necessarily involves sacrificing a portion of the data.

Results: Here we propose a flexible Bayesian model for analysis of AI, which accounts for bias and can be implemented without DNA controls. In lieu of DNA controls, this Poisson-Gamma (PG) model uses an estimate of bias from simulations. The proposed model always has a lower type I error rate compared to the binomial test. Consistent with prior studies, bias dramatically affects the type I error rate. All of the tested models are sensitive to misspecification of bias. The closer the estimate of bias is to the true underlying bias, the lower the type I error rate. Correct estimates of bias result in a level alpha test.

Conclusions: To improve the assessment of AI, some forms of systematic error (e.g., map bias) can be identified using simulation. The resulting estimates of bias can be used to correct for bias in the PG model, without data filtering. Other sources of bias (e.g., unidentified variant calls) can be easily captured by DNA controls, but are missed by common filtering approaches. Consequently, as variant identification improves, the need for DNA controls will be reduced. Filtering does not significantly improve performance and is not recommended, as information is sacrificed without a measurable gain. The PG model developed here performs well when bias is known, or slightly misspecified. The model is flexible and can accommodate differences in experimental design and bias estimation.

Show MeSH
Related in: MedlinePlus