A novel design of whole-genome microarray probes for Saccharomyces cerevisiae which minimizes cross-hybridization

Abstract

Background: Numerous DNA microarray hybridization experiments have been performed in yeast over the last years using either synthetic oligonucleotides or PCR-amplified coding sequences as probes. The design and quality of the microarray probes are of critical importance for hybridization experiments as well as subsequent analysis of the data.

Results: We present here a novel design of Saccharomyces cerevisiae microarrays based on a refined annotation of the genome and with the aim of reducing cross-hybridization between related sequences. An effort was made to design probes of similar lengths, preferably located in the 3'-end of reading frames. The sequence of each gene was compared against the entire yeast genome and optimal sub-segments giving no predicted cross-hybridization were selected. A total of 5660 novel probes (more than 97% of the yeast genes) were designed. For the remaining 143 genes, cross-hybridization was unavoidable. Using a set of 18 deletant strains, we have experimentally validated our cross-hybridization procedure. Sensitivity, reproducibility and dynamic range of these new microarrays have been measured. Based on this experience, we have written a novel program to design long oligonucleotides for microarray hybridizations of complete genome sequences.

Conclusions: A validated procedure to predict cross-hybridization in microarray probe design was defined in this work. Subsequently, a novel Saccharomyces cerevisiae microarray (which minimizes cross-hybridization) was designed and constructed. Arrays are available at Eurogentec S. A. Finally, we propose a novel design program, OliD, which allows automatic oligonucleotide design for microarrays. The OliD program is available from authors.

Figure 1

Possible types of blast alignments between a probe (query) and the S. cerevisiae genome (subject). (1) The query sequence has a continuous match in one or several different subject sequences (Type1 alignment); (2) the query sequence has discontinuous matches aigainst several regions in the same subject sequence (Type2 alignment). L_queryand L_subject are the distances (in bp) between two successive blocks of alignments in query and subject sequences, respectively.

Figure 2

Comparison of transcript level measurements between duplicated probe spots within a single array from Batch616. A scatter plot of Cy5 intensities at duplicated probe spots from a single hybridization is shown. cDNA targets were prepared from total RNA isolated from wild-type BY4742 (Cy3 labelled) and Δydr225w (Cy5 labelled) strain as described in "Methods". The 2 × limits are shown.

Figure 3

Effects of probe size, GC probe content and position relative to stop codon on signal intensities. Cy3- and Cy5-labelled cDNA targets were prepared from the total RNA isolated from the wild-type strain BY4742 and Δymr011w, respectively. Hybridization was performed as described in "Methods" with microarray from Batch616. For the Cy5 channel, normalized signals of each spot were computed as a function of probe size (panel A), GC probe content (panel B) and the distance from 3'end of the probe to stop codon of the CDS (panel C). Panel A includes PCR probes only, panels B and C include both PCR and oligonucleotide probes. Red curves represent running averages of the signal intensities.

Figure 4

Protocol strategy used in the array experiments.

Figure 5

Schematic representation of the design procedure. The main scripts (written in perl and sh shell programming languages) used in this work are in brackets []. [Newsort] allows the rearrangement of the blast output table in a positional order along the query and subject sequences. [Verifalign] detects the different types of alignments (as described in Figure 1). [Verifbarres] and [Verifbar] allow the detection of potential cross-hybridization regions in Type2 and Type1 alignments, respectively. All scripts are available upon request.

Figure 6

A view of the microarray. The array is composed of 32 grids of 420 spots each. A total of ca. 200 empty spots are distributed through the array for background controls. Probe spots are deposited in close duplicates. A set of PCR products and synthetic oligonucleotides was selected as controls. These include scorecard kits (Amersham Biosciences) in grids 4, 8, 12, 16, 20, 24, 28, and 32; serial dilutions of the signal normalization luciferase gene (row 1 in grids 3, 7, 11, 15, 19, 23, 27, 31) for which control RNA spike (Promega) can be obtained; 10 long oligonucleotides covering YKL182w (6153 bp) and 9 long oligonucleotides covering YLR310c (4767 bp) ORFs as controls for reverse transcription efficiency; 10 intergenic regions, 10 intronic sequences, and mitochondrial genes, 20 non-monotonous trinucleotide repeats (72-mer oligonucleotides) and 4 serial dilutions (in grid 1, 4, 29 and 32) of the total genomic DNA from the wild-type strain S288c; 3 E. coli genes (tufA, aceF, kdtA) as negative controls; the LexA binding domain, the LacZ 5' and 3'end regions, the Pho4 binding domain, the Gal4 binding domain, the GFP, the TAP and GST as commonly used tags or reporter genes; Leu1, His5, and Ura4 from S. pombe, the humanCBF2 andc-myc genes, and the kan^R gene as heterologous genes and markers. Printing buffer was deposited on the empty spots. The array shown results from hybridization with cDNA targets of total RNA isolated from wild-type BY4742 (Cy3-labelled) and Δydr225w (Cy5-labelled) strain.