Background: EML4-ALK gene fusions are oncogenic drivers in non-small cell lung
cancer (NSCLC), and liquid biopsies containing EML4-ALK fragments can be used to
study tumor dynamics using next-generation sequencing (NGS). However, the sensitivity
of EML4-ALK detection varies between pipelines and analysis tools.
Results: We developed an R/Bioconductor package, DNAfusion, which can be applied
to BAM files generated by commercially available NGS pipelines, such as AVENIO. Fortyeight
blood samples from a training cohort consisting of 41 stage IV EML4-ALK-positive
NSCLC patients and seven healthy controls were used to develop DNAfusion. DNAfusion
detected EML4-ALK in significantly more samples (sensitivity = 61.0%) compared
to AVENIO (sensitivity = 36.6%). The newly identified EML4-ALK-positive patients were
verified using droplet digital PCR. DNAfusion was subsequently validated in a blinded
validation cohort comprising 24 EML4-ALK-positive and 24 EML4-ALK-negative stage IV
NSCLC patients. DNAfusion detected significantly more EML4-ALK individuals in the validation
cohort (sensitivity = 62.5%) compared to AVENIO (sensitivity = 29.2%). DNAfusion
demonstrated a specificity of 100% in both the training and validation cohorts.
Conclusion: Here we present DNAfusion, which increases the sensitivity of EML4-ALK
detection in liquid biopsies and can be implemented downstream of commercially
available NGS pipelines. The simplistic method of operating the R package makes it
easy to implement in the clinical setting, enabling wider expansion of NGS-based
diagnostics.