STRmie-HD enables interruption-aware HTT repeat genotyping and somatic mosaicism profiling across sequencing platforms

STRmie-HD is a novel, alignment-free computational framework that enables high-resolution, interruption-aware genotyping and quantitative somatic mosaicism profiling of HTT repeat expansions across multiple sequencing platforms, offering superior sensitivity for characterizing Huntington's disease pathogenesis compared to conventional tools.

The Gist

The Big Picture: A New Detective for a Genetic Mystery

Imagine the human genome is a massive library of instruction manuals. In one specific book (the HTT gene), there is a paragraph that gets repeated over and over again, like a broken record. This paragraph consists of three letters: C-A-G.

In most people, this "broken record" plays a normal number of times. But in people with Huntington's Disease (HD), the record gets stuck and repeats hundreds of times. The more it repeats, the more severe the disease becomes.

For a long time, scientists have had a hard time counting these repeats accurately, especially because:

1. The repeats sometimes have "glitches" (interruptions) in the middle.
2. Different cells in the same person can have different numbers of repeats (a phenomenon called somatic mosaicism).
3. Different machines (sequencers) read the DNA in different ways, making it hard to compare results.

Enter STRmie-HD. Think of this tool as a brand-new, super-smart detective that can count these repeats, spot the glitches, and measure the chaos in the cells, no matter which machine you used to read the DNA.

---

The Problem: The "Broken Record" Analogy

Imagine you are trying to count how many times a song repeats in a very long audio file.

• The Glitch: Sometimes, the song has a tiny pause or a different note in the middle (like a C-A-A instead of C-A-G). This is called an Interruption.
• The Mosaic: Imagine you have a choir of 1,000 singers. Most sing the song 40 times. But a few singers get stuck and sing it 100 times, while others sing it only 20 times. This mix is Somatic Mosaicism.
• The Old Tools: Previous software was like a person trying to count the singers by listening to a blurry recording. They could guess the average, but they often missed the singers who were singing way too many times, or they couldn't hear the tiny "glitch" notes that change the song's meaning.

The Solution: STRmie-HD

The authors built STRmie-HD to solve these problems. Here is how it works, using our analogies:

1. It Doesn't Need a Map (Alignment-Free)

Most software tries to line up the DNA reading against a "perfect" reference map. If the DNA is too messy or too long, the map breaks.

• STRmie-HD is like a detective who doesn't need a map. It just looks at the raw audio file (the DNA read) and counts the notes directly. It can handle the "broken record" even if it's extremely long or messy.

2. It Spots the "Glitches" (Interruption Variants)

Sometimes, the repeating pattern changes slightly.

• LOI (Loss of Interruption): Imagine the song usually has a tiny pause (a glitch) that stops the repetition from getting too crazy. If that pause disappears, the song runs wild. This is very dangerous for the patient.
• DOI (Duplication of Interruption): Imagine the pause is duplicated, making the song even more complex.
• STRmie-HD is the only tool that can say, "Hey, 30% of these singers have lost the pause, and 5% have an extra pause." Old tools often missed this or just gave a vague "maybe."

3. It Measures the "Chaos" (Somatic Mosaicism)

The tool calculates two special scores:

• The Expansion Index (EI): This measures how many "wild" singers are in the choir singing way too many times.
  • Real-world finding: The tool found that the "choir" in the brain (cortex and striatum) has way more wild singers than the "choir" in the blood. This confirms that the disease is much more active in the brain than in the blood, which is a crucial insight for doctors.
• The Instability Index (II): This measures how "wobbly" the group is. Are they shifting toward singing more or fewer times?

The Proof: Did it Work?

The researchers tested STRmie-HD on four different types of "audio recordings" (datasets):

1. Illumina (Short reads): Like listening to short clips of the song. STRmie-HD was more accurate than the old tools at counting the repeats.
2. PacBio (Long reads): Like listening to the whole song in one go. STRmie-HD caught the exact number of repeats, even when other tools got confused by very long songs.
3. Synthetic Data: They made up fake songs with specific glitches to test the tool. STRmie-HD found every single glitch perfectly.
4. Nanopore (Noisy reads): This is like listening to the song in a loud, windy room. STRmie-HD still managed to count the repeats accurately, whereas other tools struggled to hear anything.

Why Does This Matter?

• Better Diagnosis: Doctors can now get a more precise count of the repeats and know exactly if dangerous "glitches" are present. This helps predict how fast the disease might progress.
• Clinical Trials: When testing new drugs, researchers need to know exactly who has the "wild" versions of the gene. STRmie-HD helps sort patients into the right groups.
• One Tool for All: Instead of needing a different software for every type of DNA machine, scientists can now use this one tool for almost everything.

The Bottom Line

STRmie-HD is a powerful new software that acts like a high-definition microscope for Huntington's disease genetics. It doesn't just count the repeats; it listens for the subtle "glitches" and measures the chaos in the cells, giving doctors and researchers a much clearer picture of the disease than ever before.

Technical Summary

1. Problem Statement

Huntington's Disease (HD) is caused by an expansion of CAG trinucleotide repeats in exon 1 of the HTT gene. While the repeat length correlates with disease onset, the clinical phenotype is heavily influenced by two complex factors that current computational tools struggle to resolve simultaneously:

• Sequence Interruptions: Specific motifs (e.g., Loss of Interruption [LOI] where a CAA triplet is lost, or Duplication of Interruption [DOI]) significantly alter disease progression and somatic instability. Existing tools often infer these indirectly or fail to quantify them at the single-read level.
• Somatic Mosaicism: The HTT repeat length varies across cells within a tissue (especially in the brain vs. blood). Standard tools often provide aggregate genotypes, failing to capture the "long right tail" of expanded alleles that drives pathology.
• Platform Limitations: Current bioinformatic solutions are often platform-specific (e.g., optimized only for Illumina or only for long-reads) and rely on reference-based alignment, which can fail for highly expanded or interrupted repeats that map poorly to the reference genome.

2. Methodology: STRmie-HD

The authors present STRmie-HD, a novel, alignment-free, de novo analytical framework designed for high-resolution genotyping of the HTT exon 1 locus.

• Core Mechanism: Instead of aligning reads to a reference genome, STRmie-HD uses a custom regular expression (regex) engine to parse individual sequencing reads. It identifies CAG and CCG motifs, calculates uninterrupted tract lengths, and detects specific interruption variants (LOI-CAA, LOI-CCA, DOI).
• Input Flexibility: It accepts raw FASTA/FASTQ files from multiple platforms:
  • Illumina (Short-read): Requires merged paired-end reads.
  • PacBio SMRT & Oxford Nanopore (Long-read): Includes a specific Region-of-Interest (ROI) parsing mode (--nanopore flag) that locates flanking anchor sequences to mitigate high error rates, extracting the repeat tract for analysis.
• Allele Detection: Generates a histogram of CAG/CCG repeat counts per sample. It employs multiple peak-detection strategies (scipy find_peaks, user-defined cutpoints, or Continuous Wavelet Transform) to identify the two primary alleles.
• Quantitative Metrics:
  • Interruption Profiling: Calculates the exact percentage of reads carrying LOI or DOI motifs.
  • Somatic Indices:
    • Instability Index (II): Measures the asymmetry of the repeat distribution around the expanded allele (bias toward expansion vs. contraction).
    • Expansion Index (EI): Quantifies the accumulation of signal beyond the expanded allele peak, serving as a proxy for the burden of somatic mosaicism.
• Output: Produces an HTML report containing allele calls, histograms, interruption percentages, and somatic indices.

3. Key Contributions

• Unified Framework: STRmie-HD is the first tool to provide a single, consistent pipeline for HTT genotyping across Illumina, PacBio, and Oxford Nanopore platforms without requiring reference alignment.
• Single-Read Interruption Resolution: Unlike tools that infer interruptions from aggregate shifts, STRmie-HD quantifies the exact fraction of reads containing specific interruption variants (LOI/DOI) at single-read resolution.
• Novel Somatic Metrics: It implements and validates the Expansion Index (EI) and Instability Index (II) for targeted NGS data, allowing for the quantitative assessment of somatic mosaicism burden.
• Dual-Tract Analysis: Simultaneously genotypes both the CAG (polyQ) and CCG (proline-rich) tracts, providing a complete molecular picture of the HTT exon 1 architecture.

4. Results

The tool was benchmarked against state-of-the-art tools (ScaleHD, RepeatDetector, TRGT, DeepRepeat, CharONT2) across four distinct datasets:

• Dataset 1 (Illumina MiSeq, n=368):
  • Accuracy: STRmie-HD achieved low Mean Absolute Error (MAE: 0.26 for normal, 0.70 for expanded) and high correlation (Spearman's $\rho$ = 0.93) with PCR/CE ground truth, outperforming ScaleHD (which struggled with large expansions) and matching RepeatDetector.
  • Interruption Detection: Correctly identified 7 orthogonally validated LOI-CAA samples (30–48% read fraction) that ScaleHD missed entirely and RepeatDetector only inferred indirectly.
• Dataset 2 (PacBio SMRT, n=22):
  • STRmie-HD showed superior accuracy for expanded alleles compared to TRGT (MAE 0.32 vs. 5.86 for expanded alleles in specific cases) and matched RepeatDetector.
• Dataset 3 (Synthetic Long-Reads, n=10):
  • Perfectly recovered ground-truth CAG counts and interruption proportions (LOI/DOI) across a spectrum of clinical scenarios.
• Dataset 4 (Oxford Nanopore, n=11):
  • Achieved high concordance with Sanger sequencing (MAE 1.27), outperforming CharONT2 in terms of diploid completeness (recovered both alleles in 100% of cases vs. 27% for CharONT2).
• Biological Validation (Somatic Mosaicism):
  • Analysis of blood vs. brain (cortex/striatum) tissues showed that the Expansion Index (EI) was significantly higher in brain tissues ($p < 0.003$), correctly reflecting the known biological reality that somatic expansion is more severe in the CNS than in peripheral blood.

5. Significance

• Clinical Relevance: By accurately detecting LOI variants (which can accelerate onset by up to 25 years) and quantifying somatic mosaicism, STRmie-HD improves risk stratification and patient selection for clinical trials.
• Research Utility: The ability to quantify the "burden" of somatic expansion (via EI) provides a new biomarker for monitoring disease progression and therapeutic efficacy in HD.
• Accessibility: The tool is alignment-free, computationally efficient, and user-friendly (CLI-based with HTML reports), making it accessible to labs with limited bioinformatics expertise.
• Extensibility: The regex-based, de novo architecture allows the framework to be adapted for other repeat expansion disorders (e.g., SCA1, Fragile X, Myotonic Dystrophy) by simply modifying the target motif.

In conclusion, STRmie-HD fills a critical gap in the HTT analysis landscape by offering a robust, platform-agnostic solution that captures the full complexity of repeat architecture, including interruptions and somatic mosaicism, which are essential for understanding HD pathogenesis.