Page 101 - EJMO-9-3

Eurasian Journal of
Medicine and Oncology Novel senescence-based melanoma risk model

as protective factors. Among these, baculoviral IAP repeat- each gene to the overall risk score is proportional to its
containing 3, C-C motif chemokine ligand 8 (CCL8), prognostic significance. The training set samples were
protein kinase C beta, suppressor of cytokine signaling classified into high-risk and low-risk groups based on
1, TNF superfamily member 14, and zeta chain of T cell the median risk scores. This stratification revealed a clear
receptor-associated protein kinase 70 were markedly linked dichotomy in patient outcomes, such as individuals in the
to infiltration of activated CD4 memory T cells, while low-risk group showed remarkably better overall survival
+
C1QA, CCL4, and CCL5 were strongly associated with compared to those in the high-risk group (p<0.001)
CD8 T cells infiltration (Figure 2D and E). In addition, (Figure 3B). This finding indicates that the risk model
+
CXC motif chemokine ligand 1, CXC motif chemokine effectively stratifies patients according to their likelihood of
ligand 5, and CXC motif chemokine receptor 2, acting as survival, suggesting its potential utility in clinical practice
risk factors (HR > 1), were significantly associated with for predicting patient outcomes. To further validate the
neutrophil infiltration, which in agreement with previous prognostic independence of the risk score, a multivariate
studies. Moreover, activation scores of key immune Cox regression analysis was performed, adjusting for key
34
hallmark pathways, including interferon (IFNs) response, clinical variables including age, gender, tumor stage (T
inflammatory response, and interleukin-6/Janus kinase/ stage), nodal involvement (N stage), and distant metastasis
signal transducer and activator of transcription 3 (IL-6/ (M stage). The results confirmed that the risk score showed
JAK/STAT3) signaling, were markedly elevated in Cluster a significant association with patient survival, suggesting its
2 compared to Cluster 1 (Figure 2F). These pathways play potential as an independent predictor of patient outcomes
crucial roles in regulating immune responses, enhancing (Figure 3C). Notably, age, advanced T4 stage, and N3 stage
anti-tumor immunity, and modulating the TME. The were also identified as independent risk factors for poor
heightened activation of these immune-related pathways prognosis, aligning with established clinical knowledge
in Cluster 2 suggests a more robust immune surveillance and further validating the model’s alignment with real-
mechanism, which may contribute to improved tumor world patient outcomes. Intriguingly, an examination
control and more favorable clinical outcomes. Collectively, of risk scores across the previously identified molecular
these findings provide strong evidence that the subtypes revealed that patients in cluster 2 – characterized
immunoactivated subtype is closely associated with better by prolonged survival and an immunoactivated phenotype
prognosis, further highlighting the potential significance – consistently exhibited lower risk scores compared to those
of immune modulation in melanoma progression and in cluster 1 (Figure 3D). The model’s ability to integrate
treatment response. molecular and clinical data into a unified risk score offers
a promising tool for risk stratification in melanoma, with
3.3. Construction of risk model based on prognostic potential applications in guiding immunotherapy decisions
senescence-related genes and identifying high-risk patients who may benefit from

After performing Lasso regression, a total of 25 genes more intensive therapeutic approaches.
were selected from the pool of prognostic senescence-
related genes to develop the risk model with multivariate 3.4. Validation of the risk model in the validation set
Cox regression (Figure 3A). Among these, ten genes were and external datasets
identified as risk factors, with forkhead box M1 emerging To confirm the accuracy of the risk model, risk scores
as the most significant contributor to poor prognosis. were measured for the validation set samples. According
Conversely, fifteen genes functioned as protective factors, to the median risk score, samples were categorized into
with inhibitor of NF-κB kinase regulatory subunit gamma high-risk and low-risk groups. The low-risk group showed
and CCL8 showing the strongest associations with favorable significantly better survival compared to the high-risk
survival outcomes. Notably, CCL8 exhibited significantly group, indicating that the risk model accurately stratifies
higher expression levels in Cluster 2, suggesting its potential patients based on their prognostic likelihood (p=0.0018,
role in shaping the immunoactivated TME. Previous Figure 4A). Next, differences in immune profiles between
correlation analysis revealed that CCL8 expression showed the two groups were evaluated using CIBERSORT analysis,
a positive correlation to the infiltration of activated CD4 followed by the Wilcoxon rank-sum test. The low-risk
+
memory T cells, reinforcing its possible involvement group exhibited significantly elevated levels of activated
in enhancing anti-tumor immune responses. For each NK cells, anti-tumor M1 macrophages, and CD8 T
+
sample, the risk score was determined as the weighted sum cells, consistent with the immune profile observed in the
of the selected gene expression, multiplying each gene’s training set (Figure 4B), indicating an immunoactivated
expression by its corresponding regression coefficient, subtype. To further assess the robustness of the risk model,
which was obtained from the multivariate Cox regression an additional SKCM dataset (GSE65904) was obtained
model. This approach ensures that the contribution of from the GEO database. After filtering out probes
17
Volume 9 Issue 3 (2025) 93 doi: 10.36922/ejmo.8574