Page 73 - TD-4-3

Tumor Discovery Highly accurate gene panels for cancer screening

4.5. Other challenges concepts of N-genes and T-genes, characterized by gene

Several other challenges remain, such as the role of low- expression intervals populated only by normal and tumor
expressed genes, the method’s performance in highly samples, respectively. The construction of perfect gene
heterogeneous tumors, and the possible impact of batch panels represents the first practical application of these
effects. In principle, our gene panels are robust against concepts, which we anticipate can be translated into
these concerns. Specifically, our N- and T-genes exhibit flexible and effective diagnostic tools.
distinct expression intervals populated only by a significant In addition, this paper presents arguments supporting
fraction of N and T samples, respectively. They are not low- the use of perfect panels in tumor taxonomy and highlights
expressed genes. Regarding batch effects, the TCGA data their gene members as candidate targets of therapeutic
used for panel discovery is largely free from such biases, as applications. Other potential applications, such as early
all samples were processed using a consistent technological diagnosis and efficacy monitoring, alongside challenges,
framework and standardized procedures. For each tissue, like technical standardization and cost considerations in
there is a single batch of normal samples and a single batch clinical implementation, are particularly important and
of tumor samples. Concerning tumor heterogeneity, because warrant further attention. Research in this direction is
the taxonomy derived from a panel is comprehensive, the currently in progress.
panel should be capable of detecting tumors regardless of
their mosaic composition or degree of heterogeneity. Acknowledgments

5. Conclusion The author, Augusto Gonzalez is grateful to Rolando Perez
for a careful reading of the manuscript. The author, Gabriel
We have shown that it is possible to construct a Gil thanks Laura Azor, Fabiana Fuentes, Jorge Mato, and
combinatorial gene panel that acts as a perfect biomarker Karen Alfaro for critical comments and suggestions.
for cancer. By monitoring the gene expression profile of
the panel members, samples can be accurately classified as Funding
either normal or tumorous. In some cases, it is possible to
classify a sample as tumorous based on the overexpression The research was supported by the Financial and
of a single gene. However, this represents just one example International Projects Office of the Ministry of Sciences,
among various panel types, all of which are highly sensitive Cuba (project PN692LH007-095).
and specific. Conflict of interest
Our study analyzed 12 cancer types from the TCGA
database, encompassing many of the most prevalent The authors declare that they have no competing interests.
cancers in the world. Panels are provided on a per- Author contributions
cancer-type basis, tailored to each specific context.
A comprehensive inventory of these panels can be found in Conceptualization: Gabriel Gil, Augusto Gonzalez
the Supplementary Information. Despite the fact that other Formal analysis: Gabriel Gil, Julio C. Drake-Pérez
panels combining classifier genes could be constructed, Investigation: All authors
these are not discussed in the present paper. Methodology: Gabriel Gil, Augusto Gonzalez
Writing–original draft: Gabriel Gil, Augusto Gonzalez
While a single gene can have sufficient discriminative
power in one tissue, other tissues require panels of up Writing–review & editing: All authors
to nine genes to achieve the same level of accuracy. Ethics approval and consent to participate
Figure S4 shows the relationship between panel length
and the distance between the centers of the normal and Not applicable.
tumor sample clusters in gene expression space. 66,67 It is
evident that the shorter inter-cluster distances correspond Consent for publication
to greater overlap between normal and tumor expression Not applicable.
profiles, complicating classification and necessitating
larger panels. The figure also suggests that the inter-cluster Availability of data
distance in gene expression space functions as a global The data used was taken from TCGA public network
tumor classifier, a factor often overlooked in tumor studies. (https://portal.gdc.cancer.gov/). Relevant software is

Our gene discovery framework extends beyond the available at the GitHub repository https://github.com/
paradigm of differential expression by introducing the gabriel-gil/GenePan.

Volume 4 Issue 3 (2025) 65 doi: 10.36922/TD025190035