A novel cost-effective diagnostic tool for detection of genetic diseases

A novel cost-effective diagnostic tool for detection of genetic diseases

Purpose: To develop a software platform for real-time automated analysis and detection of genetic diseases using Oxford Nanopore Technologies (ONT) and artificial intelligence.

Background: Genetic diagnosis of chromosomal abnormalities is done through karyotyping/FISH or chromosomal microarrays (CMA). Resolution, although sufficient for large anomalies, is limited for small chromosomal rearrangements and mosaicism. Recently, Optical Genome Mapping (OGM) showed 100% concordance with karyotype and CMA and was able to detect mosaicism; however computational analysis is challenging and cost-effectiveness needs to be demonstrated. Also, analysis methodology based on CMA, cannot efficiently detect triploidies. Oxford Nanopore Technology (ONT) uses long-read fragments facilitating identification of aneuploidies and mosaicism in a rapid and inexpensive way although bioinformatics analysis is under development.

Methods: We developed Phivea®, a software solution that analyzes the .fastq files generated by GridIONx5 (ONT) and accesses them using a shared file system. The analysis includes four phases: Quality check (read length 900-1200bp with average Phred quality score above 8); Demultiplexing (using Torchlex, a proprietary method for real-time demultiplexing ONT reads); Chromosome classification; and Genetic disease classification and analysis report generation. As proof-of-concept, we evaluated classification ability for Klinefelter syndrome (KS) using samples from KS patients, healthy donors and samples from other chromosomal abnormalities. DNA libraries were prepared and loaded on the GridIONx5. Analysis and disease classification was performed on Phivea®.

Results: Our technology can be applied directly on the stream of base-called DNA reads generated by the ONT device. It exceeds the limits of the real-time monitoring and analysis per DNA sample, which can significantly reduce the overall costs. The calculated throughput of the analysis pipeline is 4120 reads/s measured on a referent hardware architecture using thread parallelism of 10. Also, the current version of the Phivea® platform managed to correctly identify 6 different genetic disorders (Klinefelter, Turner, Down, Edwards, Patau and Prader-Willi/Angelman syndromes) with sensitivity and specificity higher than 90%. Our technology was tested and validated on a mix of real and synthetically prepared samples, but the obtained results can be directly extrapolated for analysis and detection of other genetic disorders.

Conclusions: Our solution demonstrated a novel approach for detecting genetic disorders that can reduce costs, increasing throughput and facilitating analysis. Our technology was validated on chromosomal aberrations, but results can be directly extrapolated for screening of any other genetic disorder

Click here to download the abstract