Clinical Impact, Diagnostic Performance, and Prognostic Implications of Plasma Metagenomic Next-Generation Sequencing in Solid Organ Transplant Recipients

Journal: medRxiv
Published Date:

Abstract

Introduction: Plasma metagenomic next-generation sequencing (mNGS) may detect pathogens in solid organ transplant (SOT) recipients, but optimal patient selection and result interpretation remain uncertain. Methods: Physicians reviewed SOT recipients with first-instance clinical plasma mNGS testing (Karius, Inc.) and determined consensus microbiological diagnoses, clinical impact of results, diagnostic yield, and clinical outcome. mNGS results were compared to microbiological diagnoses. A HIPAA-compliant large language model (GPT-4) was used to analyze electronic medical record (EMR) data and predict risk of infection with atypical bacteria, invasive fungi, mycobacteria, or parasites (collectively: pre-specified organisms of presumed significance, POPS) and identify patients who had positive-impact mNGS testing. Results: Of 145 SOT recipients, 119 (82.1%) had positive tests, 42 (29.0%) had [≥] 1 POPS organism, and 27 (19.1%) had [≥] 1 organism causing positive clinical impact. Positive impact was highly correlated with POPS status, with 24 (88.9%) of 27 positive-impact organisms categorized as POPS (P<0.001). GPT-4 scores accurately identified patients with POPS diagnoses (AUC 0.86), and assigned higher scores to patients with positive test impact (P=0.001). mNGS testing had highest sensitivity for atypical bacteria (82.4% sensitivity) and lower sensitivity for Aspergillus spp (53.3% sensitivity). Detection of greater numbers of organisms by mNGS was associated with increased mortality risk (odds ratio 1.32 per organism detected). Discussion: Plasma mNGS is a valuable clinical tool in SOT recipients. Positive clinical impact is associated with detection of atypical bacteria, fungi, mycobacteria, or parasites. GPT-4 analysis of EMR data identifies patients at risk of infection from these organisms and most likely to benefit from mNGS testing.

Authors

  • Spottiswoode
  • N.; Marra
  • P. S.; Lydon
  • E. C.; Chu
  • V. T.; Radakovich
  • N.; Rodriguez
  • J.; Phan
  • H. V.; Langelier
  • C. R.; Fung
  • M.