A single-sample method for normalizing and combining full-resolution copy numbers from multiple platforms, labs and analysis methods

Abstract

Motivation: The rapid expansion of whole-genome copy number (CN) studies brings a demand for increased precision and resolution of CN estimates. Recent studies have obtained CN estimates from more than one platform for the same set of samples, and it is natural to want to combine the different estimates in order to meet this demand. Estimates from different platforms show different degrees of attenuation of the true CN changes. Similar differences can be observed in CNs from the same platform run in different labs, or in the same lab, with different analytical methods. This is the reason why it is not straightforward to combine CN estimates from different sources (platforms, labs and analysis methods).

Results: We propose a single-sample multi source normalization that brings full-resolution CN estimates to the same scale across sources. The normalized CNs are such that for any underlying CN level, their mean level is the same regardless of the source, which make them better suited for being combined across sources, e.g. existing segmentation methods may be used to identify aberrant regions. We use microarray-based CN estimates from 'The Cancer Genome Atlas' (TCGA) project to illustrate and validate the method. We show that the normalized and combined data better separate two CN states at a given resolution. We conclude that it is possible to combine CNs from multiple sources such that the resolution becomes effectively larger, and when multiple platforms are combined, they also enhance the genome coverage by complementing each other in different regions.

Availability: A bounded-memory implementation is available in aroma.cn.

Fig. 1.

Full resolution and smoothed tumor/normal CNs in a 60 Mb region on Chr 3 of TCGA sample TCGA-02-0104 as measured by four different labs based on three different types of microarray SNP and CN platforms (Table 1). The full-resolution estimates are displayed as light points and the smoothed estimates, which are available at every 100 kb, are displayed as dark colored curves. For set A there are 88 000 full-resolution CNs on Chr 3, for set B there are 38 000 CNs, and for sets C and D there are 15 000 CNs (approximately).

Fig. 2.

Smoothed tumor/normal CNs before (A) and after (B) multisource normalization. The same region as in Figure 1 is depicted (with a different vertical scale).

Fig. 3.

Smoothed CNs for the six different pairs of sources. Data from all autosomal chromosomes in one individual is displayed. Each curve depicts the overall pairwise relationship between the two datasets plotted. These curves, which are used only to illustrate the relationships, are fitted using smooth splines with five degrees of freedom.

Fig. 4.

Normalized full-resolution tumor/normal CNs in the same sample and region as in Figure 1. Source D was used as the target source, which is why the estimates from that source does not change.

Fig. 5.

Smoothed normalized CNs for the six different pairs of sources. After normalization the relationship between sources is approximately linear. The sample and loci shown are as in Figure 3.

Fig. 6.

The ROC performances for detecting a CN change based on the individual (dashed gray) CNs along, the combined un-normalized CNs (dash-dotted red) and the combined normalized (solid red) CNs. (A) the results for a change point on Chr 10 in TCGA-06-0178, (B) the results for a change point on Chr 12 in TCGA-02-0026.

Fig. 7.

Segmentation of a 400 kb region on Chr 3 using CN estimates from the individual sources (upper four panels) and the combined estimates (lower panel). In addition to increase density and effective resolution, the different sources also complement each other by covering different regions. All data are normalized across sources.