Optimal Probe Length Varies for Targets with High Sequence Variation: Implications for Probe Library Design for Resequencing Highly Variable Genes

Niall Haslam, Nava Whiteford, Gerald Weber, Adam Prügel‐Bennett, Jonathan W. Essex, Cameron NeylonPLoS 2008 · DOI

Abstract

Optimal probe length and the optimal number of reference sequences used to design a probe library are highly target specific for highly variable sequencing targets. The optimum design cannot be determined simply by inspection of input sequences or of alignments but only by detailed analysis of the each specific target. For highly variable sequences, shorter probes can in some cases provide better information than longer probes. Probe library design would benefit from a general purpose tool for analysing these issues. The formal terminology developed here and the analysis approaches it is used to describe will contribute to the development of such tools.

Introduction

Sequencing by Hybridization (SBH) offers a cost-effective alternative to traditional sequencing methods. By utilizing a pre-defined library of short DNA "probes" that hybridize to an unknown target sequence, researchers can infer the target's composition. However, the application of standard "tiling arrays"—typically employing probes of 15–30 nucleotides—is severely limited when dealing with highly variable viral targets such as HIV or Influenza.

In rapidly mutating genes, a single probe often encounters multiple mutations (double mutations), resulting in false negatives or ambiguous data. While "universal" libraries containing all possible sequences of a specific length could theoretically solve this, they become physically impossible to manufacture for lengths exceeding 10 nucleotides. This study addresses the challenge of designing "generic" libraries that balance practical size with the robustness needed to capture the diversity of evolving viral drug targets.

Methodology

The researchers established a formal mathematical framework to analyze probe library design, focusing on three core components:

  1. Consensus Tiling: Libraries were constructed by sliding a window of length $l$ over a consensus sequence derived from multiple known variants, rather than a single wildtype sequence.
  2. Thermodynamic Filtering: To ensure experimental consistency, the authors utilized a "thermal equivalence parameter" ($\omega$). This index, based on nearest-neighbor interactions, allows for the creation of isothermic libraries by filtering out probes that hybridize too weakly or too strongly at a specific temperature.
  3. Variability Metrics: A "relative variability order" ($R$) was introduced to quantify population-level fluctuations in a gene, correlating sequence complexity with library performance.

Datasets

The study analyzed high-profile viral drug targets from the Los Alamos National Laboratory Sequence Databases:

  • HIV-A: Reverse transcriptase (rt) and protease (pr).
  • Hepatitis C (HCV-1a): E1, core, and ns4a genes.
  • Influenza (H3N2): np5 gene.

Results and Discussion

The analysis revealed that optimal probe length is highly variable, ranging from 12 to 19 nucleotides depending on the target.

The Shorter-is-Better Paradox

A key finding is that for extremely variable genes, shorter probes can outperform longer ones. While longer probes are generally more unique, they are significantly more susceptible to "breaking" when encountering multiple mutations. Shorter probes navigate high-variation regions more effectively, provided they maintain sufficient length to remain unique within the target genome.

Target-Specific Performance

  • Influenza np5: This target achieved approximately 90% identification with a relatively small library (~4,600 probes) at an optimal length of 13 nt.
  • HIV Reverse Transcriptase: Required a longer optimal length (16 nt) to manage sequence repetitions, achieving ~48% coverage.
  • HCV ns4a: This target proved the most challenging. With a high relative variability ($R = 0.526$), it resulted in less than 9% coverage even with optimized parameters.

Thermal Impact

The stringency of thermal filtering ($\Delta\omega$) directly influences optimal probe length. In the np5 gene, adjusting the thermal window shifted the optimal length from 12 to 15 nt, highlighting that design cannot be divorced from experimental conditions.

Conclusion

The study demonstrates that optimal probe design is a complex interplay between mutation patterns and thermodynamic constraints. For highly "chaotic" genes like HCV ns4a, reference-based libraries may fail entirely, necessitating conventional sequencing. The formal terminology and optimization logic presented here lay the groundwork for a general-purpose computational tool to automate probe library design for variable targets.

Summary updated

Abstract

Optimal probe length and the optimal number of reference sequences used to design a probe library are highly target specific for highly variable sequencing targets. The optimum design cannot be determined simply by inspection of input sequences or of alignments but only by detailed analysis of the each specific target. For highly variable sequences, shorter probes can in some cases provide better information than longer probes. Probe library design would benefit from a general purpose tool for analysing these issues. The formal terminology developed here and the analysis approaches it is used to describe will contribute to the development of such tools.
BibTeX
@inproceedings{haslam2008optimal,
  author = {Niall Haslam and Nava Whiteford and Gerald Weber and Adam Prügel‐Bennett and Jonathan W. Essex and Cameron Neylon},
  title = {{Optimal Probe Length Varies for Targets with High Sequence Variation: Implications for Probe Library Design for Resequencing Highly Variable Genes}},
  booktitle = {PLoS},
  year = {2008},
  doi = {10.1371/journal.pone.0002500},
}
Figure