In a recent study published in Science, researchers used genome-wide analysis of cell-free DNA (cfDNA) fragment end positions and surrounding sequences to detect the presence of tumour-derived DNA in blood. They developed a new metric called the information-weighted fraction of aberrant fragments (iwFAF) based on genome-wide differences in fragment positioning, weighted by fragment length and GC content. This new metric could be used in liquid biopsies for cancer diagnostics.
Analysing plasma DNA for cancer diagnostics
Plasma DNA analysis (also known as liquid biopsy) has been used to develop new diagnostic techniques in cancer. The detection and quantification of tumor-derived plasma DNA has mainly relied on analyzing somatic genetic alterations. However, there are only a few genes and genomic loci that are recurrently altered in most cancer patients. Moreover, genes commonly mutated in cancer can also be affected in non-malignant conditions, making the early detection of cancer less effective. Another approach is to profile the epigenome by analysing plasma DNA methylation, looking for a combination of both tissue-specific and cancer-specific alterations.However, this approach relies on tumour-derived DNA, which is often present in very low levels in early stages of cancer – again making early detection less effective, as well as making it necessary to obtain plasma DNA from multiple tubes of blood.
All this suggests we need better liquid biopsy biomarkers. One solution some studies have attempted is to trade depth of molecular analysis for breadth of genomic analysis, by leveraging genomic features that capture how DNA shed from different cell types is processed and fragmented in the blood. The fragmentation characteristics of cfDNA reflect the chromatin accessibility in the cells that shed DNA into plasma. In this study, researchers used an approach called genome-wide analysis of fragment ends (GALYFRE), which “aggregates genomic positioning of breakpoints across all sequenced fragments in a sample”. This means GALYFRE looks at the distribution of breaks across the entire genome, rather than focusing on specific regions or genes. By looking at the overall pattern of breaks across the genome, GALYFRE aims to identify unusual patterns or changes that may be indicative of cancer or other conditions. This approach differs from earlier studies that focused on specific fragment lengths or sequence motifs in individual cfDNA fragments.
Schematic shows how GALYFRE calculates nucleotide frequencies by measuring the frequency of each nucleotide at specific positions along the fragment ends of cfDNA.
Genomic fragment analysis
In this study, researchers analysed more than 2000 samples from patients with cancer to develop a metric based on genome-wide differences in fragment positioning, weighted by fragment length and guanine-cytosine (GC) content, which they termed information-weighted fraction of aberrant fragments (iwFAF). The researchers first measured and compared aberrant fragmentation in healthy individuals and patients with cancer, and also compared aberrant fragmentation to the fraction of tumour DNA in plasma, and evaluated differences in genomic positioning of plasma DNA fragments by measuring nucleotide frequencies at fragment ends. They found that iwFAF strongly correlated with the fraction of tumour DNA in the blood, was higher for DNA fragments carrying somatic mutations, and was higher within genomic regions affected by copy number amplifications. They then measured any potential confounding factors that could affect iwFAF measurements, and found that factors such as sample processing conditions, age groups, gender, collection tube types and more, make no significant difference to iwFAF measurements.
The researchers then developed a machine-learning model to differentiate healthy individuals from patients with cancer. This random-forest machine-learning model had an averaged area under the receiver operating characteristics curve (AUC) value of 0.91 – an AUC value of a model is a measure of its performance, specifically its ability to distinguish between true positive and true negative results. AUC ranges from 0 to 1, with a value of 0.5 indicating no discrimination between positive and negative results, and a value of 1 indicating perfect discrimination. Therefore, an AUC value of 0.91 indicates that the model has a very high ability to distinguish between plasma samples from patients with cancer and healthy individuals. The results from this study suggest that analysis of fragment ends can be a cost-effective and accessible approach for cancer detection and monitoring – especially for early-detection of cancer.
