Page 30 - GTM-3-3

Global Translational Medicine Computational advances in cancer liquid biopsy

specific information than DNA and RNA measurements, within single cells and highlight differential junction usage
helping to identify tumor origin and drug targets more across cell subpopulations.
confidently. This is an important asset concerning methods Methods are rapidly evolving. As thorough analyses
limited to DNA/RNA analysis, which sometimes show poor of cfDNA, plasma proteome, and multiomics analyses
correlation with disease outcome and response. Moreover, of CTCs increase the comprehensiveness of tumor
proteins are the principal mediators of most cellular molecular profiling, spatial omics techniques (genomics,
100
functions as well as the primary target of approved drugs transcriptomics, epigenomics, and proteomics )
103
102
101
for human diseases. Even though proteomic data holds provide complementary information by dissecting
significant potential in novel biomarker identification and solid tumor architecture and highlighting intercellular
clinical implementation, nowadays, discoveries in this interactions between tumor cells and microenvironment.
field lag behind those in genomics and transcriptomics, Still, while tools for the analysis of single-cell data are
and researchers are striving to find new, valuable protein now mature and well-established, there remains room for
biomarkers. One reason is that the proteome is far more improvement in the computational methods used for the
dynamic than the genome and transcriptome due to analysis of spatial omics datasets at single-cell or nearly
the high number of protein isoforms, which constantly single-cell resolution, particularly in reducing noise,
and rapidly change in response to internal/genetic and
94
external/environmental stimuli. Throughput represents performing deconvolution, and optimizing segmentation
95
another limitation, as the study of protein biomarkers for precise cellular characterization.
in liquid biopsies relies mostly on traditional antibody- 7. ML
based approaches or liquid-bead immunoassays, which are
unsuitable for high-plex proteomic profiling. Sensitivity ML-based, fast, reproducible, automated, and bias-free
96
is a further challenge, as most tissue-specific biomarkers enumeration of CTCs is considered a promising alternative
are present at ultra-low concentrations in blood plasma. to operator visual scoring. Given that CTCs are very rare
97
Recently, several efforts have been made to address these and difficult to detect, missing even a single CTC can lead
challenges at both the plasma and CTC levels. High to patient misclassification. However, some challenges
sensitivity, high multiplexing, speed, and automation have faced by human operators also apply to ML-based
been achieved with the NULISA assay, which allows the methods. The precise morphological characterization of
98
simultaneous measurement of hundreds of proteins in CTCs remains unknown, primarily due to their scarcity
patients’ plasma, detecting both low- and high-abundance and diverse forms, resulting in high variability and poor
proteins within the same sequencing library. Meanwhile, reproducibility in human-made decisions when it comes to
104
Wolf and colleagues presented TEMPO, an approach for CTC identification. Within clinical samples, a spectrum
integrating microvolume liquid-biopsy high-resolution of cellular states is often observed, spanning from intact to
proteomics with tissue-cell-level transcriptomics and damaged cells, with varying fluorescent intensities across
artificial intelligence (AI) to examine disease mechanisms, all channels. Consequently, establishing an exact decision
even from liquid biopsies collected from non-blood boundary for CTC identification is difficult and should be
sources. 99 regarded as a compromise between precision and recall. 105
Notably, innovative technologies are continuously Numerous ML models have been developed to
being developed, allowing simultaneous characterization characterize CTCs, 106-109 focusing on genomic features,
of the genome, transcriptome, and surface proteome of gene expression, the epigenome, or phenotypic
rare tumor cells. GoT-Splice is a cutting-edge technology information such as cell shape, size, and fluorescence.
92
for single-cell multiomics integration and represents a step Recently, pirone and colleagues favored label-free
110
forward in studying clonal mosaicism in human samples. approaches over marker-based ones like the FDA-
This innovative approach facilitates the joint profiling of approved CellSearch system, proposing an imaging data-
genotype, gene expression, protein, and aberrant splicing crunching algorithm based on cells’ 3D, quantitative, and
within individual cells, enabling researchers to compare cells biophysical properties. The authors employ tomographic
with somatic mutations to those with wild-type genotypes phase imaging technology to first distinguish CTCs from
within the same sample. By doing so, it allows the analysis white blood cells and subsequently discriminate between
of the impact of somatic mutations on transcriptional different cancer cell lines. The current well-established
and splicing phenotypes, empowering the inference of technologies for isolating CTCs in liquid biopsy samples
genotype-to-phenotype relationships. Leveraging long- are based on immunogenicity, CTC collection (positive
read sequencing of scRNA-seq, the authors developed a enrichment), removal of blood cell subpopulations
pipeline to detect and quantify alternative splicing events (negative enrichment), and enrichment based on

Volume 3 Issue 3 (2024) 7 doi: 10.36922/gtm.3063