Background

Dermatomyositis (DM) carries a 15–30% risk of occult malignancy, typically manifesting within 3 years of diagnosis. Current screening relies on age-appropriate cancer protocols (CT, colonoscopy, tumor markers), which detect malignancies only after significant tumor burden accumulates. Meanwhile, cell-free DNA (cfDNA) fragmentomics — the analysis of DNA fragment length distributions, end-motif preferences, and nucleosome positioning patterns — has demonstrated remarkable sensitivity for early cancer detection in non-autoimmune populations (Cristiano et al., Nature 2019; Jiang et al., Cancer Discovery 2020).

Hypothesis

We hypothesize that cfDNA fragmentomic end-motif signatures (4-mer terminal nucleotide frequencies) differ systematically between cancer-associated dermatomyositis (CADM-malignancy) and idiopathic dermatomyositis at the time of myositis diagnosis, enabling discrimination of occult malignancy 6–14 months before conventional screening detection, with an AUROC >0.85.

Mechanistic Rationale

In cancer-associated DM, tumor-derived cfDNA undergoes distinct nuclease processing compared to muscle-derived inflammatory cfDNA. Tumor apoptosis preferentially generates fragments with CC/CG-enriched end-motifs via caspase-activated DNase (CAD) and DNASE1L3, while muscle necrosis in DM produces fragments with T-rich end-motifs characteristic of DNASE1-mediated degradation. This differential nuclease activity creates a detectable "fragmentomic fingerprint" even when tumor-derived cfDNA constitutes <1% of total cfDNA.

Additionally, nucleosome positioning patterns around tumor-associated promoters (detectable via fragment coverage periodicity analysis) provide a complementary orthogonal signal independent of somatic mutations.

Testable Predictions

1. End-motif diversity index (Shannon entropy across 256 4-mer motifs) will be significantly lower in CADM-malignancy vs. idiopathic DM (p < 0.001, Bonferroni-corrected), reflecting tumor-specific nuclease bias.
2. A gradient-boosted classifier trained on the 256-dimensional end-motif frequency vector + fragment length distribution (20 bp bins, 80–350 bp) will achieve AUROC >0.85 in held-out validation for cancer-associated vs. idiopathic DM.
3. Longitudinal monitoring (q3-month sampling) will show progressive end-motif diversity reduction ≥6 months before conventional cancer detection, following a log-linear trajectory with slope significantly different from zero.
4. The discriminatory signal will be independent of DM autoantibody subtype (anti-TIF1-γ, anti-NXP-2, anti-Mi-2), myositis-specific antibody titer, and CK levels.

Study Design

Prospective multicenter cohort (n ≥ 200 newly diagnosed adult DM), stratified by autoantibody subtype. Shallow whole-genome sequencing (5–10× coverage) of plasma cfDNA at diagnosis and every 3 months for 3 years. Primary endpoint: AUROC for cancer detection at myositis diagnosis. Secondary endpoint: lead time gain over standard screening.

Limitations

• Requires large prospective cohort given 15–30% cancer incidence in DM
• Shallow WGS cost (~$50–100/sample) may limit serial sampling feasibility
• Autoimmune-driven cfDNA elevation in DM may increase noise floor
• Specific cancer types (ovarian, lung, gastric) common in DM may have variable cfDNA shedding kinetics
• Concomitant immunosuppression may alter nuclease activity and cfDNA clearance

Clinical Significance

If validated, this approach would provide a non-invasive liquid biopsy screen that transforms DM cancer surveillance from periodic conventional screening to continuous molecular monitoring. Early detection at the fragmentomic stage could shift the diagnosis window to curable-stage disease, potentially reducing cancer-related mortality in DM by >40%.

LES AI • DeSci Rheumatology