Pediatric leukemias: clinical validation of Oxford Nanopore-based gene expression profiling

Biography

Dr Javeria Aijaz leads the molecular pathology and biorepository sections at Indus Hospital & Health Network, Pakistan. Her research focuses on implementing nanopore sequencing for paediatric cancer diagnostics in resource-limited settings through collaborations with St. Jude Children’s Research Hospital.

Abstract

While cancer diagnostics traditionally relies on methods like flow cytometry, immunohistochemistry, karyotyping, and fluorescence in situ hybridisation (FISH), gene expression profiling (GEP) has recently emerged as a viable alternative approach. Limited data is available on the clinical validation of these approaches against standard testing. The goals of this study were to (a) ascertain the accuracy of GEP using Oxford Nanopore Technologies sequencing for pediatric acute leukemias lineage identification against standard flow cytometry (b) determine other clinical performance characteristics (e.g. precision, limit of detection). In this study, we analyzed 368 pediatric acute leukemia samples collected at Indus Hospital & Health Network (IHHN) in Karachi, Pakistan, with data analysis performed at the University of North Carolina (UNC). The specimens, comprising peripheral blood and/or bone marrow mononuclear cells, were preserved in either liquid nitrogen or Zymo DNA/RNA shield at –80°C in IHHN's pediatric biorepository. Each case included comprehensive diagnostic data from IHHN's CAP-accredited laboratory, including clinical history, complete blood count (CBC), flow cytometry, FISH, and karyotype results. Using a previously established probability threshold of >0.8 for confident classification, 95% of specimens (321/338) were correctly called. Thirty samples in total were not confidently called and interpreted as test failure with the need to reflex to flow cytometry. Of the 21 samples tested for precision in a total of 64 runs, discordance was observed in three runs. A threshold blast percentage of 20 was found to be predictive for a correct call. Technical factors such as N50, and number of aligned reads did not appear to substantially influence correct call rate.