3.1 The prognostic value IRGs of LUAD

Supplementary Table 1 displayed the ITH score of LUAD cases. We discovered that a high ITH score was associated with a high pT stage, a clinical stage, and a poor prognosis in LUAD cases after classifying them into high ITH score and low ITH scores (Supplementary Fig. 1A-1C). We found 629 DEGs mediating ITH in LUAD using the criteria of |LogFC|≥ 2 and p-value < 0.001 (Supplementary Fig. 1D). Twenty genes were identified as potential prognostic biomarkers for LUAD by using univariate cox analysis (Supplementary Fig. 1E).

3.2 A predictive IRS built by integrative machine learning algorithms

In order to create a predictive IRS, the aforementioned potential prognosis markers were fed into an integrative machine learning software, which produced 101 different types of prediction models. The best IRS prediction, as seen by Fig. 1A, was the Lasso + stepCox(both) based prognostic signature, which had the greatest average C-index of 0.80. Seven genes were used to build the Lasso + stepCox(both) based IRS, and the IRS score of LUAD could be computed using the formula below: (0.0638 × TLR7exp + 0.0123 × PRIM2exp + 0.1659 × ITGALexp + (−0.0986) × FYCO1exp + 0.1136 × CDCA3exp + (−0.0753) × CAV3exp + 0.0695 × ANK2exp). Following the division of LUAD cases into high and low IRS score groups, as illustrated in Fig. 1B-J, we discovered that LUAD patients with high IRS score had a poor clinical outcome. The AUCs of 1-, 3-, and 5-year ROC were, respectively, 0.884, 0.804, and 0.782 in the TCGA cohort; 0.839, 0.733, and 0.752 in the GSE30219 cohort; 0.850, 0.791, and 0.798 in the GSE31210 cohort; 0.792, 0.835, and 0.827 in the GSE37745 cohort; 0.933, 0.693, and 0.713 in the GSE42127 cohort; 0.881, 0.814, and 0.795 in the GSE50081 cohort; 0.854, 0.752, and 0.756 in the GSE68465 cohort; 0.844, 0.813, and 0.818 in the GSE68571 cohort; 0.799, 0.758 and 0.753 in GSE72094 cohort, respectively (Fig. 1B-J).

Fig. 1

Machine learning developed a prognostic IRS. A The C-index of 101 kinds prognostic models of TCGA and all GEO datasets. The survival curve of different IRS score groups and their corresponding ROC curve in TCGA (B), GSE30129 (C), GSE31210 (D), GSE37745 (E), GSE42127 (F), GSE50081 (G), GSE68467 (H), GSE68571 (I) and GSE72094 (J) datasets

3.3 IRS acted as prognostic indicator in LUAD

The results of the cox regression analysis were displayed in Fig. 2A, B, showing that IRS was an independent risk factor for the prognosis of LUAD patients in the TCGA and GEO cohorts. In every group, the C-index of IRS was greater than other clinical characteristics (age, gender, and clinical stage) (Fig. 2C). The C-index of IRS was shown to be greater than many prognostic markers in the TCGA cohort, despite the fact that numerous prognostic models have been constructed for LUAD (Supplementary Table 2 and Fig. 2D). In order to forecast the clinical prognosis of LUAD cases, a nomogram was also created, and excellent agreement was found between the ideal and prediction curves (Fig. 3A, B). When compared to IRS, age, gender, and clinical stage, the nomogram's AUC and predictive utility were greater (Fig. 3C, D).

Fig. 2

Evaluation of the role of IRS in LUAD patients. A, B Potential marker identified by univariate and multivariate cox regression analysis. C The C-index of IRS, age, gender and clinical stage in TCGA AND GEO datasets. D The C-index comparing the value of IRS and other 100 established signatures in evaluating the prognosis of LUAD patients in TCGA dataset

Fig. 3

Development of a predictive nomogram. A Nomogram developed based on IRS and clinical characters. B, C Calibration and ROC curve evaluated the role of nomogram in the prognosis of LUAD patients. D DCA curve suggested the good potential of the nomogram for clinical application

3.4 Significant correlation between ITH score and tumor immune microenvironment

There was a significant association found between the IRS score and the abundance of immune cells (Fig. 4A, p < 0.05). The IRS score in LUAD was inversely correlated with the number of B cells, CD8+ T cells, NK cells, and dendritic cells (Fig. 4B-E). Low IRS score were associated with increased levels of B cells, CD8+ T cells, mast cells, and TILs in LUAD patients, as Fig. 4F illustrates. Additionally, there is higher gene set score correlated with T cell co-stimulation, cytolytic activity, and immune checkpoint in LUAD patients with low IRS score (Fig. 4G, all p < 0.05). In addition, a low IRS score in LUAD was associated with a higher immunological score and ESTIMAE score (Fig. 4H, all p < 0.001).

Fig. 4

Correlation between IRS score and immune microenvironment. A The correlation between IRS and the abundance of immune cell based on seven algorithms. B–E The level of B cells, CD8+ T cells, NK cells, and dendritic cells was negatively correlated with IRS score. F, G The level of immune cells and immune-related function in different IRS score groups. H The immune score and ESTIMAE score in LUAD cases with different IRS score. *p < 0.05, **p < 0.01, ***p < 0.001

3.5 IRS acted as an indicator for predicting therapy benefits in LUAD

Higher IPS indicated better immunotherapy benefits, and immunophagoscore was a more accurate predictor of immunotherapy outcomes. [13]. Improved immunotherapy outcomes are also connected with high TMB scores [14]. Figure 5A–D shows that LUAD cases with high IRS score also had higher immunophenoscore for PD1 and CTLA4, higher TMB score, lower immune escape score, and lower ITH score. Additionally, LUAD patients with low IRS score exhibit increased expression of immunological checkpoints (Fig. 5E, p < 0.05). A higher chance of benefiting from immunotherapy was linked to high expression of HLA-related genes [15]. HLA-related genes were more prevalent in LUAD cases with poor IRS score (Fig. 5F, all p < 0.05). Therefore, LUAD with a low IRS score may benefit more from immunotherapy, and IRS may serve as an indicator of immunotherapy benefit. To confirm the findings, three more immunotherapy datasets were used. The IRS score of respondents in the IMvigor210 cohort was downregulated in comparison to non-responders, as Fig. 5G illustrates. Patients with higher IRS score showed poorer OS rate and lower responder rate. Similar outcomes were also seen in the GSE78220 and GSE91061 cohorts (Fig. 5H, I). It was also confirmed that IRS was successful in predicting the medication sensitivity in LUAD. Figure 6A, B demonstrate that LUAD patients with low IRS score had higher IC50 values for 5-Fluorouracil, Cisplatin, Docetaxel, Epirubicin, Gemcitabine, Crizotinib, Dasatinib, Erlotinib, Foretinib, and Gefitinib.

Fig. 5

IRS acted as an indicator for immunotherapy benefits in LUAD. The level of PD1&CTLA4 immunophenoscore (A), TMB score (B), immune escape score (C), ITH score (D), immune checkpoints (E) and HLA-related genes (F) in different IRS score group. G–I The overall rate and immunotherapy response rate in different IRS score group in IMvigor210, GSE91061 and GSE78220 cohort. *p < 0.05, **p < 0.01, ***p < 0.001

Fig. 6

The IC50 value of drugs in different IRS score group. The IC50 value of chemotherapy (A) and target therapy (B) in different IRS score group

3.6 The relationship between IRS and cancer characteristics in LUAD

We then performed a functional enrichment analysis to see why there were substantial differences in the clinical outcome and immunotherapy advantages amongst the various IRS score groups. A higher gene set score connected with angiogenesis, coagulation, DNA repair, glycolysis, hypoxia, IL2-STAT5 signaling, mTORC1 signaling, P53 pathway, EMT signaling, and NOTCH signaling were found in LUAD cases with higher IRS score, as Fig. 7A-J illustrated (all p < 0.05).

Fig. 7

The correlation between cancer related hallmarks and IRS in LUAD. (A-J) Low IRS score indicated a lower sore of gene sets correlated with cancer hallmark pathways