Sequence dependence of cross-hybridization on short oligo microarrays

Abstract

One of the critical problems in the short oligo microarray technology is how to deal with cross-hybridization that produces spurious data. Little is known about the details of cross-hybridization effect at molecular level. Here, we report a free energy analysis of cross-hybridization on short oligo microarrays using data from a spike-in study. Our analysis revealed that cross-hybridization on the arrays is mostly caused by oligo fragments with a run of 10-16 nt complementary to the probes. Mismatches were estimated to be energetically much more costly in cross-hybridization than that in gene-specific hybridization, implying that the sources of cross-hybridization must be very different between a PM-MM probe pair. Consequently, it is unreliable to use MM probe signal to track cross-hybridizing signal on a corresponding PM probe. Our results also showed that the oligo fragments tend to bind to the 5' ends of the probes, and are rarely seen at the 3' ends. These results are useful for microarray design and data analysis.

Figure 1

Scatter plot of signal correlation for each probe between the three replicated experiments. (A) PM probes; (B) MM probes. x-axis is the Pearson's correlation of PM probe signals between experiment batches 1 and 2 (r12); y-axis is the Pearson's correlation of PM probe signals between experiment batches 1 and 3 (r13). Red-dot spots are marked as the probes satisfying these criteria: r12 > 0, r13 > 0 and r12² + r13² > 0.81. These probes are identified as the probes with reproducible signals across three replicated experiments. Blue-cross spots are marked as the probes targeted to spike-in transcripts.

Figure 2

Distribution of SD of log-transformed probe signals for PM probes. The distribution shows a major peak and a minor peak, which represent contributions from random noise and the effects of the spike-ins, respectively. Inset shows the minor peak that is only visible after zoom-in.

Figure 3

Histogram of sizes of cross-hybridizing fragments. The fragments were collected from BLAST alignments between spike-ins and the PM probes that were affected by cross-hybridization to the spike-ins.

Figure 4

Linearity of cross-hybridization signals. Seven probes that were found to cross-hybridize to group 1 spike-ins are included here. Each line represents a probe. The x-axis shows the spike-in concentration. The y-axis shows the averaged probe signals over three replicated experiments. This figure shows that the cross-hybridization signals respond linearly to source transcripts.

Figure 5

Free-energy model fitting. (A) Nearest-neighbor stacking energy. (B) Weight factors for cross hybridization. (C) Model fitting. The model fitted binding affinities (Â) are plotted against the observed affinities (A) on a logarithmic scale.

Figure 6

Location of cross-hybridizing fragments on the probes. Each row represents a cross-hybridizing fragment from the spike-ins. For each oligo-fragment, the line represents probe (5′ end at the left); the center of a circle indicates the mid-point of oligo-fragment on the probe; and the area of the circle represents the magnitude of cross hybridization binding affinity. With a few exceptions, the oligo-fragments appear to bind stronger at the 5′ ends of the probes. A cross-hybridization oligo-fragment is shown here only if it was found to cross-hybridize to multiple probes. (A) Fragments collected from PM probes. (B) Fragments collected from MM probes.

Figure 7

Stacking energies and weight factors from PDNN model. (A) Stacking energies of NSB and GSB; (B) Weight factors of GSB and NSB. The nucleotide positions were counted from the 5′ end of the probes. The array type used here was human genome HG-U133A manufactured by Affymetrix Inc. The weight factors for NSB are higher on the left-hand side of the figure, showing a bias to the 5′ end.

Figure 8

Scatter plot of SD of log-transformed probe signal of probe pairs. x-axis is for PM probes and y-axis for MM probes. Each point represents a probe pair. Red cross represents the probes targeting spike-ins. Except for the red crosses, most of the others are close to either x-axis or y-axis, showing very discordant behavior.

Figure 9

Predicting cross-hybridization effects. (A) Distribution of free energy of binding of chMM probes and randomMM probes. The MM probes targeted to the spike-ins were excluded in this figure. chMM represents the MM probes affected by cross-hybridization. They were selected according to the following criteria: r12 > 0, r13 > 0, r12² + r13² > 0.81 and SD of log-transformed probe signal >0.3. randomMM represents the rest of MM probes. The free energy values (−G/k_BT on x-axis) were calculated according to the parameters shown in Figure 5 and using BLAST alignments of MM probe sequences with the spike-ins. (B) Scatter-plot of r12 versus r13. The data are the same as shown in Figure 1B. The probes shown in red were selected if −G/k_BT >5.5 k_BT.