The choice of the filtering method in microarrays affects the inference regarding dosage compensation of the active X-chromosome

Abstract

Background: The hypothesis of dosage compensation of genes of the X chromosome, supported by previous microarray studies, was recently challenged by RNA-sequencing data. It was suggested that microarray studies were biased toward an over-estimation of X-linked expression levels as a consequence of the filtering of genes below the detection threshold of microarrays.

Methodology/principal findings: To investigate this hypothesis, we used microarray expression data from circulating monocytes in 1,467 individuals. In total, 25,349 and 1,156 probes were unambiguously assigned to autosomes and the X chromosome, respectively. Globally, there was a clear shift of X-linked expressions toward lower levels than autosomes. We compared the ratio of expression levels of X-linked to autosomal transcripts (X∶AA) using two different filtering methods: 1. gene expressions were filtered out using a detection threshold irrespective of gene chromosomal location (the standard method in microarrays); 2. equal proportions of genes were filtered out separately on the X and on autosomes. For a wide range of filtering proportions, the X∶AA ratio estimated with the first method was not significantly different from 1, the value expected if dosage compensation was achieved, whereas it was significantly lower than 1 with the second method, leading to the rejection of the hypothesis of dosage compensation. We further showed in simulated data that the choice of the most appropriate method was dependent on biological assumptions regarding the proportion of actively expressed genes on the X chromosome comparative to the autosomes and the extent of dosage compensation.

Conclusion/significance: This study shows that the method used for filtering out lowly expressed genes in microarrays may have a major impact according to the hypothesis investigated. The hypothesis of dosage compensation of X-linked genes cannot be firmly accepted or rejected using microarray-based data.

Figure 1. Comparison of differentially expressed genes (B versus HEK cells) by RNA sequencing and microarrays.

Data are drawn from Sultan et al. and genes detected by RNA-Seq (at least five reads) but not detected by microarrays (detection score <0.95) were selected (1,640 genes in total). The plot shows log2 ratios of expression in RNA-Seq (x axis) and microarrays (y axis).

Figure 2. Box plots of the median expression levels in human monocytes according to chromosome when selecting the transcripts detected in at least 95% of individuals.

(A) Males and (B) Females.

Figure 3. Quantile functions of median expression levels of X-linked and autosomal transcripts in human monocytes.

(A) Males and (B) Females: X-linked transcripts are shown by the plain black curve, autosomal transcripts by the dashed black curve and each autosome by an individual grey curve. For a probability p (x-axis), the y-axis shows the median expression level below which p×100% of transcripts fall. The horizontal red line corresponds to the filtering performed when selecting transcripts detected in ≥95% of individuals. The vertical green line corresponds to excluding equal proportions (57.5%) of the less expressed genes on the X chromosome and on autosomes separately.

Figure 4. Comparison of expression levels between the X chromosome and autosomes in human monocytes.

(A) Males and (B) Females: The graph plots the X∶AA ratio of median expression of X-linked genes to autosomal genes according to the proportion of transcripts filtered out prior to analysis, using either a uniform threshold (red triangles) or individual thresholds on the X and on autosomes (green circles). Error bars show the 95% bootstrap confidence intervals. The horizontal dashed lines show the ratios expected if there was no dosage compensation (X∶AA = 0.5) or full compensation (X∶AA = 1). The vertical line corresponds to the proportion of genes filtered out when using a detection score ≥95%. (C) Males and (D) Females: X∶AA ratios when the X is compared to individual autosomes and the same proportion of transcripts (50%) is filtered on the X and on each autosome.

Figure 5. Expression levels of X-linked and autosomal transcripts in different human and mouse tissues.

The graph shows boxplots of expression levels either when filtering the genes according to a uniform detection threshold (detection score (DS) ≥95%) or when excluding the 50% lowest gene expressions separately on the X chromosome and on autosomes (AUT).

Figure 6. Expression levels of imprinted and non-imprinted autosomal transcripts in human moncoytes.

The graph shows boxplots of expression levels either when filtering the genes according to a uniform detection threshold (detection score (DS) ≥95%) or when excluding the 50% lowest gene expressions separately in imprinted and non-imprinted genes.

Figure 7. X∶AA ratio estimated from data simulated under different models of dosage compensation and assuming variable proportions of expressed genes on the X and on autosomal chromosomes (see legend of x axis).

In all cases, 50% of transcripts were filtered out prior to analysis, using either a uniform threshold (red triangles) or individual thresholds on the X and on autosomes (green circles).