Page 25 - Read Online

Ding et al. J Transl Genet Genom 2021;5:50-61 I http://dx.doi.org/10.20517/jtgg.2020.01 Page 59

the risk scores of metastasis generated by the iPAM classifier for validation samples was highly significant
-11
(P-value = 2.68 × 10 ). Accuracy of early metastasis prediction from the iPAM classifier was higher for
young patients (AUC = 0.87) and middle-aged patients (AUC = 0.82) than for old patients (AUC = 0.69).
The trend of differences in prediction accuracy among the age groups is consistent with the observation
that the metastasis-associated immune responses in the tumor microenvironment in young and middle-
age patients were more pronounced than in old patients. Taken together, these findings show that more
pronounced metastasis-associated immune responses in tumor microenvironment are found in young and
middle-age patients than in old patients, potentially explaining the difference in accuracy of metastasis
prediction among the three age groups.

This study has limitations. Although the sample sets included 58 young patients (≤ 50 years) for identifying
DEGs and 197 young patients (≤ 55 years) for iPAM classifier development and validation, the sample
size of young patients is modest. Second, we identified genes that serve as prognostic biomarkers and also
show functional relevance to cancer progression based on IPA analysis; however, performing functional
studies on the role of the 36 iPAM genes on cancer progression was beyond the scope of this study.
Moreover, as previously reported, the Decipher GRID data sets may over-represent patients with adverse
[31]
clinicopathologic features .
We identified an iPAM classifier for prediction of early metastasis; the prediction accuracy of the iPAM
classifier was higher for young (≤ 55 years) and middle-aged patients (56-70 years) than for old patients
(> 70 years). We also provided evidence that this age-related difference in prediction accuracy can be
explained by differential immune responses to metastasis development among the three age groups.

DECLARATIONS
Acknowledgements
The authors thank Charles Warden for downloading the decipher GRID data sets from the Decipher
Biosciences.

Authors’ Contributions
Conception and design: Neuhausen SL, Ding YC
Development of methodology: Neuhausen SL, Ding YC
Acquisition of data: Wu H, Davicioni E, Karnes RJ, Klein EA, Den RB
Analysis and interpretation of data: Neuhausen SL, Ding YC
Writing, review, and/or revision of the manuscript: Neuhausen SL, Ding YC
Administrative, technical, or material support: Steele L
Study supervision: Neuhausen SL

Availability of data and materials
Data will be available upon request.

Financial support
This project was supported by the Morris and Horowitz Families Professorship (SLN). Research reported
in this publication included work performed in the Integrative Genomics and Pathology, Cores supported
by the National Cancer Institute of the National Institutes of Health under award number P30CA033572.
The content is solely the responsibility of the authors and does not necessarily represent the official views of
the National Institutes of Health.

Conflicts of interest
All authors declared that there are no conflicts of interest.