Frugal alignment-free identification of FLT3-internal tandem duplications with FiLT3r

Abstract

Background Internal tandem duplications in the FLT3 gene, termed FLT3-ITDs, are useful molecular markers in acute myeloid leukemia (AML) for patient risk stratification and follow-up. FLT3-ITDs are increasingly screened through high-throughput sequencing (HTS) raising the need for robust and efficient algorithms. We developed a new algorithm, which performs no alignment and uses little resources, to identify and quantify FLT3-ITDs in HTS data. Results Our algorithm (FiLT3r) focuses on the k-mers from reads covering FLT3 exons 14 and 15. We show that those k-mers bring enough information to accurately detect, determine the length and quantify FLT3-ITD duplications. We compare the performances of FiLT3r to state-of-the-art alternatives and to fragment analysis, the gold standard method, on a cohort of 185 AML patients sequenced with capture-based HTS. On this dataset FiLT3r is more precise (no false positive nor false negative) than the other software evaluated. We also assess the software on public RNA-Seq data, which confirms the previous results and shows that FiLT3r requires little resources compared to other software. Conclusion FiLT3r is a free software available at https://gitlab.univ-lille.fr/filt3r/filt3r. The repository also contains a Snakefile to reproduce our experiments. We show that FiLT3r detects FLT3-ITDs better than other software while using less memory and time.

UBTF tandem duplications define a distinct subtype of adult de novo acute myeloid leukemia

“…( E ) BM blast infiltration, ( F ) WBC count and ( G ) BM morphology classification according to FLT3-ITD status in UBTF -TD AML. The FiLT3r algorithm was used for detection and precise quantification of FTL3 -ITDs [ 23 ]. H BM morphology in patient #L171D0708G ( FLT3 -ITD with VAF < 5%) and ( I ) patient #L151D4757X ( FLT3 -ITD negative) showing characteristic dyserythropoiesis and blasts with rare Auer rods (red arrow).…”

Ex vivo drug response profiling for response and outcome prediction in hematologic malignancies: the prospective non-interventional SMARTrial

“…The no-control option of the Roddy SNVCallingWorkflow plugin 25 (version 2.1.1) and Roddy IndelCallingWorkflow plugin 25 (version 3.1.1) was used for the calling, annotation and filtering described above. FiLT3r (version b44c21f) was used to call the ITD in the FLT3 gene with default parameters, except the reference sequence was updated to exon 14 and 15 regions in the GRCh37 coordinates (13:28607897-28608566) 26 .…”

Random forest classifiers trained on simulated data enable accurate short read-based genotyping of structural variants in the alpha globin region at Chr16p13.3

“…These methods have been highly successful at calling variants throughout the genome, but because they are trained on genome-wide data and rely on accurately-aligned reads to make their predictions, they may not perform as well in repetitive regions. In addition, alignment-free methods for genotyping medically relevant variants have been developed (13)(14)(15), but so far these have been more applicable to calling smaller variants.…”