Page 136 - GPD-3-4

Gene & Protein in Disease A pyroptosis-related gene signature in myeloma

two distinct clusters (k = 2) by CDF values (Figure 2B-D). a risk model was constructed with 9 genes including
Furthermore, the first group of patients demonstrated CASP3, CHMP2A, CHMP3, CHMP6, GZMB, CASP8,
a survival advantage over the second group, as shown in NOD2, PLCG1, and FOXO3, adhering to the minimum
Figure 2E, highlighting the significant prognostic impact criteria (Figure 3A and 3B). This model effectively could
of the preliminary PRGs. In addition, comparisons of stratify the MM patients into two risk groups based on
clinical features such as tumor stage, ethnicity, and gender the median risk score (Figure 3C), with the high-risk
revealed that there were no significant differences between group showing higher mortality and shorter survival
the two clusters (Figure 2F). These results therefore suggest time (Figure 3D). Further univariate and multivariate
that these PRGs could be used to classify MM patients. Cox regression analyses established the prognostic risk
score as an independent predictor of patient prognosis
3.3. Creation of a distinct prognostic risk model (Figure 3E and 3F), with P-values below 0.001 and hazard
within the training set ratios exceeding 1. The heatmap analysis further indicated
To further examine the classification of MM patients significant differences in the gene expression profiles, of
based on the 20 PRGs as shown in Figure 2A, 842 MM which PLCG1, FOXO3, CHMP2A, CHMP3, and NOD2
patients from the TCGA-MMRF CoMMpass project were were notably downregulated in the high-risk group, while
randomly divided into the training cohort (including other genes including CASP3, CHMP6, GZMB, and CASP8
442 patients) and the test cohort (including 440 patients), were upregulated (Figure 3G).
and there were no significant differences in the general MM patients in the high-risk group showed a shorter
information between these cohorts (Table 3). Next, to total survival rate as analyzed by the Kaplan–Meier (K–M)
determine whether the 20 prognostic genes could be test, regardless of treatment (Figure 4A). The nine-gene
used to create a model for predicting prognostic risk, we risk model’s effectiveness confirmed the predicted patient
conducted a stepwise Lasso regression of the 20 PRGs survival outcomes by the values of area under the curve
within the training set, and the results generated a formula (AUC, Figure 4B). AUC measures the overall performance:
to calculate a risk score: risk score = (1.255 * CASP3 exp.) an AUC of 0.5 indicates no discrimination, while 1.0
+ (−2.177 * CHMP2A exp.) + (−1.864 * CHMP3 exp.) + signifies perfect subgroup discrimination. Finally, PCA
(1.002 * CHMP6 exp.) + (0.486 * GZMB exp.) + (1.172 * plots demonstrated a clear distinction between the two
CASP8 exp.) + (−0.554 * NOD2 exp.) + (−1.131 * PLCG1 risk groups based on 9 PRGs; however, the difference was
exp.) + (−1.536 * FOXO3 exp.). Using this formula, not as evident when considering all genes or only PRGs
(Figure 4C-E). Therefore, an independent prognostic
Table 3. Characteristics of multiple myeloma patients in two model with a 9-gene set was constructed in the training
cohorts cohort.
Training Test P-value 3.4. Confirmatory assessment of prognostic
cohort cohort signature’s robustness
Gender 0.749
The above study has established a 9-PRG signature
Male 230 262 gene model in the training set. To validate these 9 PRGs
Female 172 178 in the test cohort of 440 patients, both univariate and
Race 0.916 multivariate Cox regression analyses were performed,
White 261 301 and both analyses confirmed that the risk score functions
Asian 7 6 as an independent prognostic factor, with P-values below
Black or African American 51 61 0.001 and hazard ratios exceeding 1 (Figure 5A and 5B).
Consistent with the finding in the training set, the high-
Unknown 83 72 risk group also exhibited a higher mortality rate and a
Vital status 0.919 similar gene expression pattern for the signature genes
Alive 330 321 in the test cohort (Figure 5C-E). The K–M survival
Dead 72 119 curves also showed significant differences in the survival
Stage 0.841 probability (Figure 5F). Furthermore, the ROC curve
Stage I 141 140 analysis also found that the risk signature maintained
Stage II 140 153 strong prognostic accuracy (Figure 5G). The PCA analysis
echoed the findings from the training cohort, effectively
Stage III 112 135 distinguishing patients from among the two risk groups
Unknown 9 12 in the test cohort (Figure 5H). Overall, these results

Volume 3 Issue 4 (2024) 7 doi: 10.36922/gpd.4534