Tumor and non-tumor tissues release cellular material including DNA fragments in the blood. cfDNA carries several attributes that are informative about the source tissue contexts, such that systematic analysis of those properties can provide further information about the disease potential and tissue of origin. First, a tumor genome has 10³−10⁵ somatic mutations, only a small minority of which are classic oncogenic mutations, while the others are passenger mutations − which are typically classified as the variants of unknown significance (VUS). Since cancer is a disease marked by clonal growth, the somatic mutations of malignant origin are likely to be present in clonally amplified cell populations (Podlaha et al., 2012); therefore unlike the baseline somatic mutations from other, non-tumor tissues, the cancer-associated mutations are likely to be present at a relatively higher allele frequency in cfDNA. Somatic mosaicism (Moss et al., 2018) i.e., presence of clonal growth in healthy somatic tissues, low tumor DNA fraction in cfDNA, as well as the low depth of coverage of cfDNA sequencing could potentially complicate the inferences. Nonetheless, the VUS vastly outnumber the oncogenic mutations, such that a joint analysis of allele frequency distributions of the VUS and oncogenic driver mutations from cfDNA sequencing data can provide increased statistical power to predict the presence of clonally amplified cell populations carrying the observed set of oncogenic mutations in the patient’s blood.

Second, the somatic mutations typically carry mutational signatures of tissue-dependent carcinogenic exposure and DNA repair defects. For instance, DNA derived from lung tissues e.g. in the liquid biopsy fraction typically carries mutational signatures of smoking and other pollutants (Abbosh et al., 2017). Since a liquid biopsy sample carries an admixture of DNA fragments derived from potential tumor tissues as well as hematopoietic and other non-malignant tissues (Hu et al., 2020), which are present even in healthy individuals, an analysis of the mutational signatures in cfDNA can help identify the presence of nucleic acid contents from unexpected tissue types (Gerstung et al., 2020). In addition, cfDNA fragments in the blood typically are 120−220 bp, or multiples thereof, with a maximum peak at 167 bp − which corresponds to the length of DNA wrapped around a single nucleosome in the source tissue, and therefore the genome-wide landscape of cfDNA fragments provides information about epigenome of the cell of origin.

Third, although cancer progression is a gradual process, often progressing through localized dysplasia and hyperplasia stages for years before neoplastic transformation, not all clonal growths progress to malignant disease (Podlaha et al., 2012; Gerstung et al., 2020; Singh et al., 2020). The neoplastic transformation typically involves overcoming replicative senescence and genomic instability, accompanied by invasive clonal growth. Genomic instability leads to distinct mutational signatures in somatic mutations, which are uncommon in normal tissues. For instance, the mutational signature SBS3, SBS8, and SBS40, which are associated with replication stress and genomic instability (Maley et al., 2017; Alexandrov et al., 2020; Gerstung et al., 2020) − rarely occur in normal or benign tissues, but are widely present in cancer genomes. Collectively, the presence of oncogenic mutations, as well as the somatic mutations with distinct mutational signatures and inferred chromatin makeups in cfDNA can help identify potentially malignant clonal growth in a narrow set of probable tissue types with significant contributions in the cfDNA.

Fourth, it is increasingly becoming evident that the tumor microenvironment does not only include tumor, immune, and stromal cells, but also diverse microbial entities including bacteria, fungi, and viruses (Poore et al., 2020). While earlier works identified tumor-associated microbiome predominantly in selected tumor types, recent studies suggest that tumor-associated microbiome might be present in most cancer types (Sr et al., 2020). Furthermore, microbial DNA from the tumor microenvironment could be detected alongside DNA from host tissues during unbiased total cfDNA sequencing from a liquid biopsy. Since different cancer tissue types have distinct microbial makeups (Sr et al., 2020), annotation of microbial DNA could provide additional clues about the potential sources of genetic material in blood. Indeed, it seems possible to correctly predict tumor types from microbial signatures in blood alone (Sr et al., 2020). Therefore, total DNA sequencing from blood can provide orthogonal metagenomic signatures characteristics of different cancer types, adding yet another line of evidence for the detection of the disease.

While the above attributes could be derived from the standard cfDNA sequencing, additional insights could be gained from profiling microRNA and DNA methylation states, and other biomarkers, which typically require dedicated assays. Taken together, in addition to the detection of oncogenic mutations the emerging molecular signatures in cfDNA sequencing (Figure 1) can provide multi-faceted insights into tumor etiology and disease status, and guide imaging or molecular marker-based targeted follow-ups and treatment options.

A number of genomic approaches and computational methods have been developed for sensitive detection of somatic mutations and other molecular signatures from cfDNA sequencing. Recent developments have increased the yield and quality of cfDNA through a rational design (De, 2011), and it has been possible to isolate tumor-derived DNA from diverse types of body fluids including sputum, saliva, and urine to inform about cancer progression (Ignatiadis et al., 2021; Jj et al., 2020). In parallel, using specialized genomic and computational approaches it is now possible to detect somatic mutations at an ultra-low allele frequency (Mw et al., 2016; Sr et al., 2020). Moreover, integrative analysis of driver and passenger somatic mutations from cfDNA has been able to infer clonal and subclonal mutations, and also clonal architecture in the primary tumor for some patients (Abbosh et al., 2017)–which opens up the possibility of tracking clonal dynamics in vivo in real-time. Elegant computational analyses of cfDNA fragment patterns in patient samples have provided quantitative estimates of the likely tissue of origin of cfDNA (Schwarzenbach et al., 2011). More recently, Poore et al. used an innovative approach to identify cancer types based on microbial signatures in blood (Cc et al., 2020), which has significance for cancer diagnostics. Collectively, these and other advances provide a rich toolbox to track cancer progression non-invasively from blood, even at an early stage of the disease and recurrence.