Abstract
The relative positioning of DNA probes on microarray slides and source plates introduces subtle but significant correlations between pairs of genes. Careful consideration of this spatial artifact is important for analysis of microarray expression data. It is particularly relevant to recent microarray analyses that suggest that co-expressed genes cluster along chromosomes or are spaced by multiples of a fixed number of genes along the chromosome.
Introduction
In the early 2000s, microarray technology revolutionized biology by allowing researchers to measure the expression of thousands of genes simultaneously. Naturally, scientists began looking for patterns, specifically "co-expression"—when groups of genes turn on or off together. Several high-profile studies suggested that genes located near each other on a chromosome, or even spaced at regular intervals, were co-expressed, implying a high-level "chromosomal architecture" of gene regulation.
However, researchers suspected a "ghost in the machine." Because microarray probes are often printed onto slides in an order that mirrors their chromosomal sequence, any technical noise affecting a specific area of the slide (like a smudge, a scratch, or variation in printing tip pressure) would create a fake correlation between those neighboring genes. This study aims to decouple true biological signal from these spatial artifacts.
Methodology
The researchers analyzed the relationship between gene expression correlation and two types of distance:
- Chromosomal Distance: How far apart genes are on the DNA strand (measured in Open Reading Frames, or ORFs).
- Chip Distance: How far apart the physical DNA probes are on the microarray slide.
They used Fourier transforms to identify periodicities (repeating patterns) in the data and compared "intra-chromosomal" pairs (genes on the same chromosome) with "inter-chromosomal" pairs (genes on different chromosomes). Since genes on different chromosomes shouldn't share biological regulation based on proximity, any correlation between them that scaled with physical distance on the chip had to be an artifact.
The Local Chip Artifact
A universal trend was identified where the closer two probes are on a slide, the higher their average correlation. This "neighborhood noise" can be caused by physical imperfections on the slide or the chemistry of the hybridization process.
As shown in the figure above, average gene pair correlation is highest at short chip distances and decays as the distance increases, regardless of whether the genes are biologically related (intra-chromosomal) or not (inter-chromosomal). This decay proves that physical proximity on the slide induces spurious correlation.
Debunking Genomic Periodicities
In yeast, previous studies noted co-expression periodicities of 2 genes and 22 genes. This study demonstrates these are almost certainly artifacts of the printing blocks and source plate layouts, not a feature of the yeast genome.
In yeast datasets, the authors identified a long-range periodicity of ~22 genes (~42 Kb) and a short-range one of 2 genes (~4 Kb). Both were mapped back to the physical layout of the microarray printing process. For example, in C. elegans data, the co-expression curve showed distinct "nodes" corresponding exactly to the 8x4 block spacing of the microarray printer.
Conclusion
While these artifacts are prevalent, completely removing them without losing real biological signal is difficult. The study suggests a "filtering" approach: excluding specific experiments or slide regions that show obvious physical defects. This work serves as a foundational warning for the field: before claiming a discovery about the spatial organization of the genome, one must first rule out the spatial organization of the laboratory equipment.
Summary updated