Skip to main content
Download PDF

A gene-based score for the risk stratification of stage IA lung adenocarcinoma

Respiratory Research volume 25, Article number: 18 (2024) Cite this article

Abstract

Objective

We aim to molecularly stratify stage IA lung adenocarcinoma (LUAD) for precision medicine.

Methods

Twelve multi-institution datasets (837 cases of IA) were used to classify the high- and low-risk types (based on survival status within 5 years), and the biological differences were compared. Then, a gene-based classifying score (IA score) was trained, tested and validated by several machine learning methods. Furthermore, we estimated the significance of the IA score in the prognostic assessment, chemotherapy prediction and risk stratification of stage IA LUAD. We also developed an R package for the clinical application. The SEER database (15708 IA samples) and TCGA Pan-Cancer (1881 stage I samples) database were used to verify clinical significance.

Results

Compared with the low-risk group, the high-risk group of stage IA LUAD has obvious enrichment of the malignant pathway and more driver mutations and copy number variations. The effect of the IA score on the classification of high- and low-risk stage IA LUAD was much better than that of classical clinicopathological factors (training set: AUC = 0.9, validation set: AUC = 0.7). The IA score can significantly predict the prognosis of stage IA LUAD and has a prognostic effect for stage I pancancer. The IA score can effectively predict chemotherapy sensitivity and occult metastasis or invasion in stage IA LUAD. The R package IAExpSuv has a good risk probability prediction effect for both groups and single stages of IA LUAD.

Conclusions

The IA score can effectively stratify the risk of stage IA LUAD, offering good assistance in precision medicine.

Introduction

With the increasing prevalence of low-dose spiral CT in lung cancer screening, the detection rate of stage IA lung adenocarcinoma (LUAD) characterized by small lung nodules is increasing, but the specific treatment for this disease still faces large challenges [1]. Standard treatment (lobectomy plus complete lymph node dissection without adjuvant therapy) has a good treatment effect for most stage IA LUAD patients but may also cause overtreatment for some low-risk populations and undertreatment for high-risk populations [2,3,4,5]. How to effectively classify the high- and low-risk subtypes of stage IA LUAD is an urgent problem that needs to be solved clinically. The risk classification of stage IA LUAD based on pathological morphology and radiological characteristics shows that these characteristics are good indicators [4,5,6,7,8]. However, screening for high-risk groups still remains insufficient, and the clinical application still faces doubts and inconveniences. The massive gene expression characteristics brought by the progress of omics technology can assist greatly in understanding carcinogenesis and the precise treatment of cancers [9]. In this study, we established a gene-based classifying score (IA score) and R package (IAExpSuv) for the risk stratification of stage IA LUAD that demonstrated good clinical application value.

Methods

Transcriptomic and clinical information

The RNASeq, DNA copy number, somatic mutation data and BiospecimenClinicalData of the TCGA were downloaded through the R package TCGA-assembler [10, 11]. Clinical information was acquired from the TCGA Pan-Cancer Clinical Data Resource (TCGA-CDR) [12]. The transcriptome data and clinical information of GSE13213, GSE26939, GS30219, GSE31210, GSE41217, GSE42127, GSE50081, GSE63459, GSE68465, GSE68571, and GSE72094 were all downloaded by the R package GEO Query [13]. The transcriptome data and clinical information of early-stage LUAD (carcinoma in situ, AIS; microinvasive adenocarcinoma, MIA; invasive adenocarcinoma, IAC) of our center have been deposited in the Genome Sequence Archive in National Genomics Data Center, China National Center for Bioinformation/Beijing Institute of Genomics, Chinese Academy of Sciences (GSA-Human: HRA005169). Clinical data in the SEER database were retrieved via SEER*Stat version 8.3.9.2 software for external validation [14]. Detailed information about Platform or Methods, Normalization, Histological type, Clinical stage and etc. is provided in Additional file 1: Table S1. Original data can be found in https://0-www-ncbi-nlm-nih-gov.brum.beds.ac.uk/geo/, https://portal.gdc.cancer.gov/, https://ngdc.cncb.ac.cn/gsa-human, and https://seer.cancer.gov/.

Study design and bioinformatic analysis

Our workflow is shown in Fig. 1. Since metastasis and recurrence of non-small-cell lung cancer usually occur within 5 years after treatment, patients who are still alive for more than 5 years are often regarded as clinically cured [15,16,17]. Therefore, we used the 5-year survival rate as the criterion for identifying high-risk or low-risk subtypes of stage IA LUAD. Then 12 datasets of 837 patients with stage IA LUAD were integrated and three canonical batch effect removal methods in gene expression matrix: z-scaling (Standardization), removeBatchEffect (limma package) and Combat (sva package) were used to normalize the transcriptome data for batch effect removal [18, 19]. UMAP (UMAP package) was used to evaluate the effect of batch removal for its excellent performance in reflecting the global structure of data and improving the operation speed [20, 21]. Next, the biological differences of high-risk and low-risk subtypes of stage IA LUAD were analyzed, where Gene set enrichment analysis (GSEA) was used to analyze different pathways (ClusterProfiler package) [22] and the TCGAmutations package and maftools package were used to analyze mutation status [23]. Then, the gene-based risk score (IA score) was established, tested and validated using k-fold cross validation by machine learning (K-nearest neighbor (knn), naive Bayes (nb), neural network (nnet), partial least squares (pls), support vector machine (svm), stochastic gradient boosting (gbm), boosting (C5.0), extreme gradient boosting (xgbTree), random forest (rf), bagged classification and regression tree (treebag) and stacking were used for machine learning of transcriptome data (caret, e1071 and h2o packages) [24, 25]). Least absolute shrinkage and selection operator (LASSO)—Logistic regression were used for dimension reduction, score establishment, risk factor analysis and probability prediction (glmnet package) [26]). Furthermore, the role of the IA score in prognostic assessment, chemotherapy prediction (The Connectivity Map database was used for drug sensitivity screening [27, 28]), identification of occult metastasis and invasion of stage IA LUAD was evaluated. Finally, we developed the R package IAExpSuv based on the IA score for risk probability prediction in groups or for single stages of IA LUAD, where the GitHub platform and devtools packages were used for R package development and storage. A total of 15,708 patients with stage IA LUAD from the SEER database were used to evaluate the classical clinical parameters, and 1881 patients with stage I cancers from the TCGA-pan database were used to evaluate the significance of this score for pancancer (The number of patient samples is based on the overall survival (OS) analysis statistics).

Fig. 1
figure 1

Study design. TCGA The Cancer Genome Atlas, LUAD lung adenocarcinoma, SEER Surveillance Epidemiology and End Results, GSEA Gene Set Enrichment Analysis, CNV copy number variation, LASSO least absolute shrinkage and selection operator

Full size image

Statistical analysis

Kaplan‒Meier curves were applied to display survival probability, the log-rank test was used for statistical analysis, and Cox analysis was used to identify the risk factors for survival (survival and survminer package). The Benjamini‒Hochberg (BH) method was used for p value correction for multigroup comparisons. Receiver operating characteristic (ROC) curves were used for diagnostic analysis (pROC package). Decision curve analysis was used to evaluate the clinical benefit (dcurves and rmda packages) [29]. The function relweights was applied to evaluate the relative importance of model variables.The Kruskal‒Wallis test (function: kruskal.test) was executed for differential gene analysis under the conditions of FDR < 0.05 and abs (log2(fold change [median])) > 0.3. Nomograms were used to visualize the probability prediction of the risk scores, and bootstrapping with 1000 resamples was used for testing (functions: nomogram and calibrate). The ggplot2 package was used for graphing. R language was adopted for the above operations, with a two-sided p value (or adj. p value) < 0.05 considered statistically significant [30].

Results

Biological differences

We divided 837 patients with stage IA LUAD in 12 datasets into a high-risk group (H) (161 patients who died within 5 years), a low-risk group (L) (289 patients who survived more than 5 years), and an unknown group (U) (387 patients with an unknown 5-year survival status) (Fig. 1). We first analyzed the biological differences between H and L. The batch effect refers to the nonbiological, technological or experimental variation introduced in the experimental process, which often affects the combined processing of multiple omics datasets. The selection of the batch effect removal method mainly considers different data characteristics, eliminates technological and experimental variation, and prevents excessive adjustment of data, which could hide real biological variations and signals present in the original data [18, 31]. We selected three batch effect removal methods commonly used in gene expression profiling: z-scaling (Standardization), removeBatchEffect (R package: limma) and Combat (R package: sva). We found that z-scaling performs well in removing batch effects and preserving inherent biological differences, while the other two methods are less effective (Additional file 8: Figure S1). We ranked genes by fold change of median (H vs. L) and performed GSEA. We found that H showed more enrichment of malignant pathways (hallmark gene sets, KEGG, reactome) (Fig. 2A). The TCGA-LUAD dataset possessed data on driver mutation and copy number status, and in this dataset, we found that H had more cancer driver mutations (Fig. 2B) and showed more copy number variations (CNVs) than L (Fig. 2C).

Fig. 2
figure 2

Biological differences in high- and low-risk stage IA lung adenocarcinoma. A Different biological pathway enrichment for high-risk and low-risk stage IA LUAD by GSEA in three database (Hallmark gene sets [left], KEGG [middle], and reactome [right]) for 12 datasets merging; B Canonical driver pathway mutations in high-risk (left) and low-risk (right) stage IA LUAD in TCGA-LUAD dataset; C Copy number variations in high-risk and low-risk stage IA LUAD in TCGA-LUAD dataset. LUAD lung adenocarcinoma, GSEA Gene Set Enrichment Analysis, TCGA The Cancer Genome Atlas

Full size image

Machine learning to establish the risk score

We applied GSE30219, which had relatively more samples (both H and L), as the external validation set and the remaining 386 patients (11 datasets) as the machine learning population (Fig. 1). We first adopted k-fold cross validation (k = 5). We divided the 386 samples into five parts and identified the risk-related genes in the training set (1: k-1) by logistic regression (H vs. L) (Fig. 3A). We conducted a sanity check of general information, including survival status (high-risk vs. low-risk), sex ratio (female vs. male), age: median (min, max) and smoking status (yes vs. no) (notably, some patients had unclear smoking history) in 5 different populations (Additional file 2: Table S2). We found that the distribution of characteristics in the 5 populations was almost evenly balanced, and the classification was reasonable. The prediction effect of corresponding risk-related genes (1: k-1) in the test set (k) was evaluated by five machine learning algorithms (non-ensemble model) and six ensemble models (Fig. 1). The area under the curve (AUC) was used to assess the prediction effect. We found that the average prediction in the test set was 0.7 (0.658 for non-ensemble algorithms and 0.696 for ensemble algorithms), mainly because of the inherent limitation of the data (small sample size of the high-risk IA type) (Fig. 3B). We took the intersection of the risk-related genes (398) in the five trials (Fig. 3C) and further filtered genes by LASSO regression and finally established the risk score (IA score) based on 64 genes in the above 386 samples by LASSO-Logistic regression (Fig. 3D). The prediction performance of the IA score in the training set (11 datasets not including GSE30219) reached 0.9 and reached the upper limit of the mean prediction (0.7) in the external validation set (GSE30219) (Fig. 3E). Compared with clinicopathological factors, the IA score showed far superior discrimination power via relative weight analysis and decision curve analysis (12 datasets) (Fig. 3F, G). Detailed information on the constituent genes related to the IA score, their biological significance, population distribution and even the detailed calculation formula of the IA score are shown in Additional file 3: Table S3, Additional file 4: Table S4, Additional file 9: Figure S2, Additional file 12: Supplementary Text. We also set up a scoring system to identify high and low risk subgroup in the stage IA patients based on feature screening of gene clusters and compared it with the IA score model; detailed information on this model is also provided in Additional file 5: Table S5, Additional file 10: Figure S3, Additional file 12: Supplementary Text.

Fig. 3
figure 3

Machine learning to determine risk score for discriminating high-and low-risk stage IA lung adenocarcinoma. A Risk-related genes by logistic regression (high-risk, H vs. low-risk, L) in each training set (k[1: k-1]; k = 1,2,3,4,5); B Prediction effect (classifying H and L) of risk-related gene (1: k-1) in test set (k) by five machine learning algorithms (left) and six ensemble models (right); C Intersection of risk-related genes in five trials: OR > 1 (left) and OR < 1 (right); D LASSO regression to filtrate intersection genes for establish risk score (IA score) for 386 samples (11 datasets not including GSE30219); E Prediction effect of IA score (classifying H and L) in training set (11 datasets not including GSE30219) (left) and external validation set (GSE30219) (right); F Relative weight of IA score and classical demographic factors (left) and driver mutations (right) for risk classifying; G Discrimination power of IA score and classical demographic factors (left) and driver mutations (right) for risk classifying in decision curve analysis. OR Odds Ratio, LASSO least absolute shrinkage and selection operator

Full size image

Survival prediction

Taking the median IA score as the cut off, we evaluated the prognostic effect of the IA score for LUAD. We found that the IA score could significantly predict the OS and disease-free survival (DFS)/progression-free survival (PFS) of overall LUAD; subgroup analysis found that the IA score could clearly predict the prognosis in stage IA and IB-II patients, and the prediction effect for stage IA was better than that for stage IB-II (OS, HR[IA]: 0.178 vs. HR[IB-II]: 0.771; DFS, HR[IA]: 0.366 vs. HR[IB-II]: 0.761) (Fig. 4A, B). However, the prediction effect for stage III and IV patients was not good (Fig. 4A, B). We combined the IA score with common classical demographic factors and driver mutations to identify independent prognostic factors for stage IA LUAD. We found that advanced age, male sex, smoking history, TP53 Mut, KRAS Mut, EGFR wild-type, and IA score were factors influencing OS in univariate Cox analysis, but only IA score, male sex, and TP53 Mut were independent factors influencing OS in multivariate Cox analysis (Additional file 6: Table S6). For DFS, advanced age, smoking history, EGFR wild-type, and IA score were influencing factors in univariate Cox analysis, but only IA score was an independent influencing factor in multivariate Cox analysis (Additional file 7: Table S7). Furthermore, we demonstrated that the prediction effect of the IA score on survival was far better than that of classical pathological factors and driver mutations by decision curve analysis (Fig. 4C, D). In the TCGA-pan cancer database, we found that the IA score still had a satisfactory predictive effect for stage I cancers (Fig. 4E).

Fig. 4
figure 4

Prognostic effect of IA score in stage IA lung adenocarcinoma. OS (A) and DFS/PFS (B) curves for two categories of patients for whole LUAD and its different stages (IA, IB-II, III, and IV) according to the value of IA score (median value as the cutoff); Efficacies of IA score and classical demographic factors (left) and driver mutations (right) for predicting OS (C) and DFS (D) in decision curve analysis; E survival curves for two categories of patients for stage I cancers according to the value of IA score (median value as the cutoff). OS overall survival, DFS disease-free survival, PFS progression-free survival, LUAD lung adenocarcinoma, PFI progression-free interval, DFI disease-free interval, DSS disease-specific survival

Full size image

Risk stratification

An important purpose of risk stratification for stage IA LUAD is to identify high-risk subgroups that may require more aggressive treatment, such as postoperative adjuvant chemotherapy. We first analyzed the similarity of prognosis and gene expression between known chemotherapy-sensitive groups (high-risk stage IB and stage II populations) and stage IA subgroups classified by IA score and classical staging groups to evaluate the efficacy. We found that although subtypes of stage IA LUAD in the 8th TNM staging (IA1/2/3) have different prognoses, the high-risk subtype (IA3) still had superior survival compared to stage IB-II disease (Additional file 11: Figure S4A). However, high-risk IA subtypes based on IA score (IA H, lower IA score) demonstrated similar prognosis with stage IB-II cancer, while IA H and IB-II diseases showed poorer prognosis than the low-risk IA subtype based on IA score (IA L, higher IA score) (Fig. 5A). Furthermore, differential gene analysis showed that IA H and IB-II disease had almost the same gene expression characteristics, while IA H and IB-II tumors showed large discrepancies in their gene expression profiles compared to IA L (Fig. 5B, C).

Fig. 5
figure 5

Significance of IA score in risk stratification of stage IA lung adenocarcinoma. A OS (left) and DFS (right) curves between two risk types of stage IA LUAD patients divided by median of IA score (IA H and IA L; IA score is a prognostic protective score) and stage IB-II LUAD patients; B volcano plots for DEGs between IA L and IA H (left), between IA L and IB-II (middle), and between IA-H and IB-II (right); C Down (left) and Up (right) DEGs between IA L and IA H, between IA L and IB-II, and between IA-H and IB-II; D similarities between query signature (IA H vs. IA L) and touchstone signatures (compounds inhibiting certain pathways in lung cancer cell lines) in Connective Map database; E OS (left) and DSS (right) curves among IA1, IA2, IA3 and T1N1 LUAD patients in SEER database; F OS (left) and DSS (right) curves among IA1, IA2, IA3 and T2aN0 LUAD patients in SEER database; G OS (left) and DFS (right) curves between two risk types of stage IA LUAD patients divided by median of IA score (N0 H and N0 L) and stage T1N1 LUAD patients; H OS (left) and DFS (right) curves between two risk types of stage IA LUAD patients divided by median of IA score (T1 H and T1 L) and stage T2, T3–4 LUAD patients. LUAD lung adenocarcinoma, OS overall survival, DFS disease-free survival, DSS disease-specific survival, DEGs differentially expressed genes, SEER Surveillance Epidemiology and End Results

Full size image

Furthermore, we explored whether people in the special group showed differences in chemotherapy sensitivity by comparing prognosis between the chemotherapy group and the nonchemotherapy group, thus directly evaluating the relationship between classical classification group, IA score classification group and chemotherapy sensitivity. We found that chemotherapy had a significant survival disadvantage for stage IA LUAD, even for high-risk subtypes (IA3) in the 8th TNM staging (Additional file 11: Figure S4B–D); while this survival disadvantage was not seen for high-risk subtypes classified by IA score (IA H), it was observed for low-risk subtypes classified by IA score (IA L) (Additional file 11: Figure S4E, F). Because of the small number of IA patients receiving chemotherapy, we further evaluated whether the IA score could indicate chemotherapy sensitivity for early LUAD (IB-II population). Similarly, chemotherapy had a survival disadvantage for the unscreened IB-II population; the survival disadvantage of chemotherapy was not seen for the IB-II high-risk population classified by IA score (IB-II H) but was presented for low-risk subtypes classified by IA score (IB-II L) (Additional file 11: Figure S4G–J).

Next, based on differential gene expression analysis of the high- and low-risk IA populations, we used the Connectivity Map database (a database that enables us to predict drugs whose related genes are highly correlated with diseases and their possible molecular mechanisms of action by comparing gene expression profile data to establish the association between genes, diseases and drugs [27, 28]) to find chemotherapeutic drugs that may be effective for high-risk IA populations. Based on the analysis results, we further demonstrated that the high-risk IA subgroup was sensitive to drugs related to classical tumor driver pathways, such as the p53 pathway and receptor tyrosine kinase (RTK) pathway (Fig. 5D).

A possible explanation for the presence of a high-risk subtype of stage IA LUAD is that the tumor has metastases or invasion that are more extensive than current modalities can detect (occult metastasis or invasion), i.e., there is an “underestimation” of both N and T status with current TNM staging [2]. We found that the high-risk subtype according to the 8th TNM staging (IA3) had a more significant survival advantage than both T1N1 and T2aN0; that is, the classification of the high-risk subtype of IA in the 8th TNM staging did not overlap with the existing TNM staging in the SEER database (Fig. 5E, F). We found that the survival of the high-risk subtype of stage IA LUAD classified by IA score (N0 H) was consistent with that of T1N1, reflecting the possible discriminative function of this IA score for occult lymph node metastasis (Fig. 5G). Similarly, the high-risk subtype of stage IA LUAD classified by IA score (T1 H) had consistent survival with T2N0. This strongly suggests the possible role of the IA score in discriminating occult invasion (the actual extent of tumor invasion beyond the T1 scope defined in current detection methods) (Fig. 5H).

R package IAExpSuv

Through the logistic model, we established the survival risk model (death within 5 years) of stage IA LUAD based on the IA score, and the prediction efficiency of the model was good in cross-validation by bootstrapping with 1000 resamples (Fig. 6A, B). Since the IA score is based on a subset of Z-transformed samples, we explored how effective the risk prediction was for an individual patient. We first standardized the gene expression profiles of individual patients in two ways (MAD [median absolute deviation]: (y-median(y))/MAD (y); UQ [upper quartile]: (y)/quantile (y, 0.75); y, gene expression value) and then calculated the IA score and evaluated its risk prediction effect. After Z-transformation of MAD/UQ-standardized patients from the built-in dataset (845 IA patients with Z-transformation [MAD/UQ-Z] and without Z-transformation [MAD/UQ-Zr]), the IA score was calculated, and its risk prediction effect was evaluated. It was found that MAD/UQ-Z had a better prediction performance (AUC: 0.7, also reached the upper limit of the mean prediction in Figs. 3B, 6C. Hereby, we developed the R package IAExpSuv to predict the survival risk of IA patients, with the package including two functions, where function IAprem was for a group of patients (Z transformation, IA score transformed, and the logistic probability calculated), and function IApres was for the risk prediction of a single patient (self-normalization, Z-transformation with the built-in dataset, IA score transformed and logistic probability calculated) (Figs. 6D, 7).

Fig. 6
figure 6

Establishing R package (IAExpSuv) for predicting survival risk probability. A A nomogram for predicting survival risk (five years) probability by logistic analysis (risk probability could be easily estimated by projecting total score calculated by IA score value to the lower point scale); B the calibration curves for the nomogram (Ideal prediction: 45-degree diagonal lines; Apparent prediction: entail cohorts; Bias-corrected prediction: bootstrapping for 1000 repetitions); C predicting efficacies (AUC) of IA score for an individual patient calculated by different procedures; D basic principle and download method for the R package IAExpSuv (Functions: IAprem and IApres) to predict the survival risk of stage IA LUAD patients. AUC area under the curve, LUAD lung adenocarcinoma

Full size image
Fig. 7
figure 7

Central Picture. The IA score and R package (IAExpSuv) can effectively stratify the risk of stage IA LUAD

Full size image

Discussion

Stage IA LUAD is a clinically early malignancy (T ≤ 3 cm N0 M0) and always possesses a favorable prognosis [32]. The LCSG821 study established anatomical lobectomy as the standard treatment for stage IA LUAD [33, 34]. Many clinical trials have proven that stage IA LUAD does not benefit from postoperative adjuvant chemotherapy [35, 36]. Therefore, the National Comprehensive Cancer Network (NCCN) guidelines recommend lobectomy plus systemic lymph node dissection without postoperative adjuvant chemotherapy as the standard of care for stage IA LUAD [32]. However, stage IA LUAD still has biological and prognostic heterogeneity, and its appropriate therapeutic strategy and postoperative management remain uncertain. Previous studies predicted the prognosis and optimized treatment of stage IA LUAD based on tumor size, radiological characteristics and pathological features. For example, the death risk increases with the tumor diameter for early LUAD [37]. The prognosis of stage IA LUAD containing ground-glass components is far better than that of pure solid tumors [5]. JCOG0804 confirmed that stage IA LUAD with T ≤ 2 cm and a consolidation tumor ratio (CTR) ≤ 0.25 could benefit from sublobular treatment (mainly wedge shaped), while JCOG0802 proved that segmental resection was better for stage IA LUAD with T ≤ 2 cm and a CTR > 0.5 [3, 6]. Stage I LUAD with lepidic growth has an excellent prognosis, while micropapilla, solid components, lymphovascular invasion, and spread through air spaces often predict a poor prognosis and benefits from adjuvant chemotherapy [4, 7, 8, 38]. Therefore, in the 2021 WHO histological classification, carcinoma in situ (AIS) with pure lepidic growth and lesions less than 3 cm were defined as precursor lesions of cancer, and LUAD with mainly lepidic growth, and lesions less than 3 cm with the presence of invasive components less than 5 mm were defined as microinvasion (T1mi) [32, 39]. Furthermore, invasive stage IA LUAD (invasive component > 5 mm) was further subdivided into T1a (< 1 cm), T1b (1 cm–2 cm), and T1c (2 cm–3 cm) in the 8th TNM staging [32].

However, the previous classification is still controversial. First, we found that the prognosis of the higher risk type (IA3) defined in the 8th TNM staging was still better than that of stage IB-II disease; thus, this classification cannot suggest directly whether to chemotherapy is needed. In addition, although existing studies believe that the maximum diameter of solid components can better distinguish prognosis than the maximum diameter of whole nodules, some studies believe that lung nodules containing ground-glass components belong to a special type of lung cancer and that size, proportion of solid components and maximum diameter of nodules cannot be used as prognostic risk factors [40,41,42]. Furthermore, it is difficult and subjective to distinguish histological morphology and pathological types. Moreover, there are often mixed pathological patterns with large variation, and the relationship of the proportion of different histological components and the extent of invasion with prognosis is still controversial. Therefore, there are still certain flaws in the current risk stratification of stage IA LUAD that thus limit precise clinical treatment.

We stratified the risk of stage IA LUAD from the perspective of gene expression characteristics. The IA score and application R package (IAExpSuv) have the following three advantages. First, this classification is more intuitive. The fundamental reason for the prognostic heterogeneity of stage IA LUAD is the heterogeneity of biological characteristics, and the gene expression characteristics largely dictate biological behavior. Therefore, our risk classification based directly on the differential gene expression characteristics of high- and low-risk populations is more thorough. The IA score can classify a high-risk type with a similar prognosis to that of stage IB-II disease, indicating that the IA score can predict stage improvement. The prognosis of the subgroup with high-risk stage IA LUAD (T1N0M0) according to IA score and of T1N1 disease (T1 stage disease with lymph node metastasis) was similar, suggesting that these patients may have occult lymph node metastasis. The prognosis of the subgroup with high-risk stage IA LUAD (T1N0M0) according to IA score and of T2N0 disease (T2 disease without lymph node metastasis) was similar, suggesting that tumor invasion and expansion in these patients may exceed that typically indicated by the T1 category. These findings suggest that the IA score can identify occult metastases and invasion (lymph node metastases and invasion not detected by current detection methods, i.e., underestimated N and T stages). High-risk stage IB and stage II patients can benefit from postoperative adjuvant therapy [43, 44]. Patients with N1 lymph node metastasis or T2 stage typically benefit from adjuvant chemotherapy [35, 45,46,47]. Therefore, it is reasonable to speculate that high-risk subtypes of stage IA LUAD classified by IA score present more aggressive biological characteristics and a poorer prognosis, indicating that these patients can benefit from adjuvant chemotherapy. We also further found that adjuvant chemotherapy is harmful for low-risk subgroups of stage IA LUAD patients classified by IA score, and this conclusion can also be extended to stage IB-II, suggesting that IA score can be used as an index for the classification of low-risk subtypes of early-stage LUAD. In addition, we found that for low-risk subtypes of stage IA LUAD patients classified by IA-score, the median survival time was 175 months for OS, with a five-year OS rate of 0.917, and the median DFS time was 175 months, with a five-year DFS rate of 0.789. The classic clinical trial with OS as the main endpoint (JCOG0802) indicated that low-risk subtypes of stage IA LUAD (T ≤ 2 cm and CTR > 0.5) benefiting from sublobectomy had a 5-year OS rate of 0.943, and another canonical clinical trial with DFS as the main endpoint (CALGB140503) indicated that low-risk subtypes of stage IA LUAD (T ≤ 2 cm) benefiting from sublobectomy had a 5-year DFS rate of 0.636 [3, 48]. The low-risk subtype of stage IA LUAD classified by IA score is similar to the existing low-risk subtype of stage IA LUAD that benefits from sublobectomy in terms of both OS and DFS. Therefore, the low-risk subtype of stage IA LUAD classified by IA score may benefit from sublobectomy. Second, we used 12 different datasets for training, testing and validation of the scoring system, so the IA score had good universality. Third, gene expression profiling only requires a small sample size, and the IA score calculation is intuitive and simple, so its clinical application would be acceptable.

In addition, we further analyzed the biological significance of the IA score and its possible therapeutic targets. Through gene enrichment analysis, we found that the 64 genes in the IA score model are enriched in signaling pathways such as proliferation and immunity, and malignant proliferation and immune microenvironment are key links in the malignant progression of early-stage LUAD [49,50,51,52,53,54], reflecting the biological rationality of IA score. We further screened key genes by analyzing the expression of these genes in early-stage LUAD tissues. We found that 13 genes showed consistency between risk classification and the expression distribution of AIS-MIA-IAC (eight genes, OR > 1, risk genes or oncogenes, sustained upward trend; five, OR < 1, protective gene or tumor suppressor gene, continuous downward trend). Most of these genes also play an important role in the evaluation of clinical features and malignant biological behavior of lung cancer. The details regarding the biological significance of these genes are shown in Additional file 12: Supplementary Text.

This study has some limitations. First, due to the sample size and quality of the public datasets (it is difficult to collect complete prognostic data of stage IA LUAD), as well as the inherent difficulties (patients with stage IA LUAD have a better prognosis, and it is difficult to identify those who died within five years), the efficacy of the IA score in the validation set could only reach 0.7. Additionally, owing to the lack of detailed data on chemotherapy (regimens and performance status of patients) and the small proportion of high-risk groups (sensitivity to chemotherapy) for early-stage LUAD (IA or IB-II), we could not directly identify IA patients who would benefit from chemotherapy through IA score, as we could only clearly and directly determine patients who should not receive chemotherapy. In addition, joint analysis of radiologic characteristics, pathological features and gene expression features cannot be achieved with public datasets. Additionally, the transition of the IA score from histological findings to blood test results remains to be verified.

Conclusions

In summary, based on the biological differences between high-risk and low-risk patients with stage IA LUAD, we established the IA score and its R package IAExpSuv, which can be effectively used for prognostic classification and risk stratification to help provide precise treatment for stage IA LUAD.

Availability of data and materials

Original data can be found in https://0-www-ncbi-nlm-nih-gov.brum.beds.ac.uk/geo/, https://portal.gdc.cancer.gov/, https://ngdc.cncb.ac.cn/gsa-human, and https://seer.cancer.gov/.

Abbreviations

LUAD:

Lung adenocarcinoma

OS:

Overall survival

SEER:

Surveillance Epidemiology and End Results

TCGA:

The Cancer Genome Atlas

GSEA:

Gene Set Enrichment Analysis

CNV:

Copy number variation

gbm:

Stochastic Gradient Boosting

rf:

Random Forest

knn:

K-Nearest Neighbor

nb:

Naive Bayes

nnet:

Neural Network

pls:

Partial Least Squares

svm:

Support Vector Machine

treebag:

Bagged Classification And Regression Tree

xgbTree:

Extreme Gradient Boosting

LASSO:

Least absolute shrinkage and selection operator

BH:

Benjamini–Hochberg

ROC:

Receiver operating characteristic

AUC:

Area under the curve

DFS:

Disease-free survival

DSS:

Disease-specific survival

DEGs:

Differentially expressed genes

RTK:

Receptor tyrosine kinase

MAD:

Median absolute deviation

UQ:

Upper quartile

NCCN:

National Comprehensive Cancer Network

CDF:

Cumulative distribution function

AIS:

Carcinoma in situ

MIA:

Microinvasive adenocarcinoma

IAC:

Invasive adenocarcinoma

References

  1. Baldwin DR, Callister ME. The British Thoracic Society guidelines on the investigation and management of pulmonary nodules. Thorax. 2015;70:794–8.

    Article  PubMed  Google Scholar 

  2. Wang L, Jiang W, Zhan C, Shi Y, Zhang Y, Lin Z, Yuan Y, Wang Q. Lymph node metastasis in clinical stage IA peripheral lung cancer. Lung Cancer. 2015;90:41–6.

    Article  PubMed  Google Scholar 

  3. Saji H, Okada M, Tsuboi M, Nakajima R, Suzuki K, Aokage K, Aoki T, Okami J, Yoshino I, Ito H, et al. Segmentectomy versus lobectomy in small-sized peripheral non-small-cell lung cancer (JCOG0802/WJOG4607L): a multicentre, open-label, phase 3, randomised, controlled, non-inferiority trial. Lancet. 2022;399:1607–17.

    Article  CAS  PubMed  Google Scholar 

  4. Wang C, Yang J, Lu M. Micropapillary predominant lung adenocarcinoma in stage IA benefits from adjuvant chemotherapy. Ann Surg Oncol. 2020;27:2051–60.

    Article  PubMed  Google Scholar 

  5. Hattori A, Suzuki K, Takamochi K, Wakabayashi M, Aokage K, Saji H, Watanabe SI. Prognostic impact of a ground-glass opacity component in clinical stage IA non-small cell lung cancer. J Thorac Cardiovasc Surg. 2021;161:1469–80.

    Article  PubMed  Google Scholar 

  6. Suzuki K, Watanabe SI, Wakabayashi M, Saji H, Aokage K, Moriya Y, Yoshino I, Tsuboi M, Nakamura S, Nakamura K, et al. A single-arm study of sublobar resection for ground-glass opacity dominant peripheral lung cancer. J Thorac Cardiovasc Surg. 2022;163:289-301.e282.

    Article  PubMed  Google Scholar 

  7. Liu C, Wang LC, Chen HS, Yeh YC, Hsu PK, Huang CS, Hsieh CC, Hsu HS. Outcomes of patients with different lepidic percentage and tumor size of stage I lung adenocarcinoma. Thorac Cancer. 2022;13:2005–13.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  8. Tsuchiya T, Akamine S, Muraoka M, Kamohara R, Tsuji K, Urabe S, Honda S, Yamasaki N. Stage IA non-small cell lung cancer: vessel invasion is a poor prognostic factor and a new target of adjuvant chemotherapy. Lung Cancer. 2007;56:341–8.

    Article  PubMed  Google Scholar 

  9. Hoadley KA, Yau C, Hinoue T, Wolf DM, Lazar AJ, Drill E, Shen R, Taylor AM, Cherniack AD, Thorsson V, et al. Cell-of-origin patterns dominate the molecular classification of 10,000 tumors from 33 types of cancer. Cell. 2018;173:291-304.e296.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  10. Wei L, Jin Z, Yang S, Xu Y, Zhu Y, Ji Y. TCGA-assembler 2: software pipeline for retrieval and processing of TCGA/CPTAC data. Bioinformatics. 2018;34:1615–7.

    Article  CAS  PubMed  Google Scholar 

  11. Zhu Y, Qiu P, Ji Y. TCGA-assembler: open-source software for retrieving and processing TCGA data. Nat Methods. 2014;11:599–600.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  12. Liu J, Lichtenberg T, Hoadley KA, Poisson LM, Lazar AJ, Cherniack AD, Kovatich AJ, Benz CC, Levine DA, Lee AV, et al. An integrated TCGA pan-cancer clinical data resource to drive high-quality survival outcome analytics. Cell. 2018;173:400-416.e411.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  13. Davis S, Meltzer PS. GEOquery: a bridge between the Gene Expression Omnibus (GEO) and BioConductor. Bioinformatics. 2007;23:1846–7.

    Article  PubMed  Google Scholar 

  14. Doll KM, Rademaker A, Sosa JA. Practical guide to surgical data sets: surveillance, epidemiology, and end results (SEER) database. JAMA Surg. 2018;153:588–9.

    Article  PubMed  Google Scholar 

  15. Postmus PE, Kerr KM, Oudkerk M, Senan S, Waller DA, Vansteenkiste J, Escriu C, Peters S. Early and locally advanced non-small-cell lung cancer (NSCLC): ESMO Clinical Practice Guidelines for diagnosis, treatment and follow-up. Ann Oncol. 2017;28:iv1–21.

    Article  CAS  PubMed  Google Scholar 

  16. Watanabe K, Tsuboi M, Sakamaki K, Nishii T, Yamamoto T, Nagashima T, Ando K, Ishikawa Y, Woo T, Adachi H, et al. Postoperative follow-up strategy based on recurrence dynamics for non-small-cell lung cancer. Eur J Cardiothorac Surg. 2016;49:1624–31.

    Article  PubMed  Google Scholar 

  17. Demicheli R, Fornili M, Ambrogi F, Higgins K, Boyd JA, Biganzoli E, Kelsey CR. Recurrence dynamics for non-small-cell lung cancer: effect of surgery on the development of metastases. J Thorac Oncol. 2012;7:723–30.

    Article  PubMed  Google Scholar 

  18. Luo J, Schumacher M, Scherer A, Sanoudou D, Megherbi D, Davison T, Shi T, Tong W, Shi L, Hong H, et al. A comparison of batch effect removal methods for enhancement of prediction performance using MAQC-II microarray gene expression data. Pharmacogenomics J. 2010;10:278–91.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  19. Ritchie ME, Phipson B, Wu D, Hu Y, Law CW, Shi W, Smyth GK. limma powers differential expression analyses for RNA-sequencing and microarray studies. Nucleic Acids Res. 2015;43: e47.

    Article  PubMed  PubMed Central  Google Scholar 

  20. Diaz-Papkovich A, Anderson-Trocmé L, Gravel S. A review of UMAP in population genetics. J Hum Genet. 2021;66:85–91.

    Article  PubMed  Google Scholar 

  21. Dadu A, Satone VK, Kaur R, Koretsky MJ, Iwaki H, Qi YA, Ramos DM, Avants B, Hesterman J, Gunn R, et al. Application of Aligned-UMAP to longitudinal biomedical studies. Patterns (N Y). 2023;4: 100741.

    Article  CAS  PubMed  Google Scholar 

  22. Yu G, Wang LG, Han Y, He QY. clusterProfiler: an R package for comparing biological themes among gene clusters. OMICS. 2012;16:284–7.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  23. Mayakonda A, Lin DC, Assenov Y, Plass C, Koeffler HP. Maftools: efficient and comprehensive analysis of somatic variants in cancer. Genome Res. 2018;28:1747–56.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  24. Deo RC. Machine learning in medicine. Circulation. 2015;132:1920–30.

    Article  PubMed  PubMed Central  Google Scholar 

  25. Greener JG, Kandathil SM, Moffat L, Jones DT. A guide to machine learning for biologists. Nat Rev Mol Cell Biol. 2022;23:40–55.

    Article  CAS  PubMed  Google Scholar 

  26. Tibshirani R. Regression shrinkage and selection via the lasso: a retrospective. J R Stat Soc Ser B Stat Methodol. 2011;73:273–82.

    Article  Google Scholar 

  27. Lamb J. The Connectivity Map: a new tool for biomedical research. Nat Rev Cancer. 2007;7:54–60.

    Article  CAS  PubMed  Google Scholar 

  28. Subramanian A, Narayan R, Corsello SM, Peck DD, Natoli TE, Lu X, Gould J, Davis JF, Tubelli AA, Asiedu JK, et al. A next generation connectivity map: L1000 platform and the first 1,000,000 profiles. Cell. 2017;171:1437-1452.e1417.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  29. Vickers AJ, Elkin EB. Decision curve analysis: a novel method for evaluating prediction models. Med Decis Making. 2006;26:565–74.

    Article  PubMed  PubMed Central  Google Scholar 

  30. Chan BKC. Data analysis using R programming. Adv Exp Med Biol. 2018;1082:47–122.

    Article  PubMed  Google Scholar 

  31. Lin S, Lin Y, Nery JR, Urich MA, Breschi A, Davis CA, Dobin A, Zaleski C, Beer MA, Chapman WC, et al. Comparison of the transcriptional landscapes between human and mouse tissues. Proc Natl Acad Sci U S A. 2014;111:17224–9.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  32. Ettinger DS, Wood DE, Aisner DL, Akerley W, Bauman JR, Bharat A, Bruno DS, Chang JY, Chirieac LR, D’Amico TA, et al. NCCN guidelines insights: non-small cell lung cancer, version 2.2021. J Natl Compr Canc Netw. 2021;19:254–66.

    Article  CAS  PubMed  Google Scholar 

  33. Ginsberg RJ, Rubinstein LV. Randomized trial of lobectomy versus limited resection for T1 N0 non-small cell lung cancer. Lung Cancer Study Group. Ann Thorac Surg. 1995;60:615–22 (discussion 622-613).

    Article  CAS  PubMed  Google Scholar 

  34. Veluswamy RR, Ezer N, Mhango G, Goodman E, Bonomi M, Neugut AI, Swanson S, Powell CA, Beasley MB, Wisnivesky JP. Limited resection versus lobectomy for older patients with early-stage lung cancer: impact of histology. J Clin Oncol. 2015;33:3447–53.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  35. Pignon JP, Tribodet H, Scagliotti GV, Douillard JY, Shepherd FA, Stephens RJ, Dunant A, Torri V, Rosell R, Seymour L, et al. Lung adjuvant cisplatin evaluation: a pooled analysis by the LACE Collaborative Group. J Clin Oncol. 2008;26:3552–9.

    Article  PubMed  Google Scholar 

  36. Kato H, Ichinose Y, Ohta M, Hata E, Tsubota N, Tada H, Watanabe Y, Wada H, Tsuboi M, Hamajima N, Ohta M. A randomized trial of adjuvant chemotherapy with uracil-tegafur for adenocarcinoma of the lung. N Engl J Med. 2004;350:1713–21.

    Article  CAS  PubMed  Google Scholar 

  37. Ost D, Goldberg J, Rolnitzky L, Rom WN. Survival after surgery in stage IA and IB non-small cell lung cancer. Am J Respir Crit Care Med. 2008;177:516–23.

    Article  PubMed  Google Scholar 

  38. Chen D, Wang X, Zhang F, Han R, Ding Q, Xu X, Shu J, Ye F, Shi L, Mao Y, et al. Could tumor spread through air spaces benefit from adjuvant chemotherapy in stage I lung adenocarcinoma? A multi-institutional study. Ther Adv Med Oncol. 2020;12:1758835920978147.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  39. Nicholson AG, Tsao MS, Beasley MB, Borczuk AC, Brambilla E, Cooper WA, Dacic S, Jain D, Kerr KM, Lantuejoul S, et al. The 2021 WHO classification of lung tumors: impact of advances since 2015. J Thorac Oncol. 2021;17:362–87.

    Article  PubMed  Google Scholar 

  40. Travis WD, Asamura H, Bankier AA, Beasley MB, Detterbeck F, Flieder DB, Goo JM, MacMahon H, Naidich D, Nicholson AG, et al. The IASLC Lung Cancer Staging Project: proposals for coding T categories for subsolid nodules and assessment of tumor size in part-solid tumors in the forthcoming eighth edition of the TNM classification of lung cancer. J Thorac Oncol. 2016;11:1204–23.

    Article  PubMed  Google Scholar 

  41. Hattori A, Matsunaga T, Takamochi K, Oh S, Suzuki K. Neither maximum tumor size nor solid component size is prognostic in part-solid lung cancer: impact of tumor size should be applied exclusively to solid lung cancer. Ann Thorac Surg. 2016;102:407–15.

    Article  PubMed  Google Scholar 

  42. Ye T, Deng L, Wang S, Xiang J, Zhang Y, Hu H, Sun Y, Li Y, Shen L, Xie L, et al. Lung adenocarcinomas manifesting as radiological part-solid nodules define a special clinical subtype. J Thorac Oncol. 2019;14:617–27.

    Article  PubMed  Google Scholar 

  43. Strauss GM, Herndon JE 2nd, Maddaus MA, Johnstone DW, Johnson EA, Harpole DH, Gillenwater HH, Watson DM, Sugarbaker DJ, Schilsky RL, et al. Adjuvant paclitaxel plus carboplatin compared with observation in stage IB non-small-cell lung cancer: CALGB 9633 with the Cancer and Leukemia Group B, Radiation Therapy Oncology Group, and North Central Cancer Treatment Group Study Groups. J Clin Oncol. 2008;26:5043–51.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  44. Butts CA, Ding K, Seymour L, Twumasi-Ankrah P, Graham B, Gandara D, Johnson DH, Kesler KA, Green M, Vincent M, et al. Randomized phase III trial of vinorelbine plus cisplatin compared with observation in completely resected stage IB and II non-small-cell lung cancer: updated survival analysis of JBR-10. J Clin Oncol. 2010;28:29–34.

    Article  CAS  PubMed  Google Scholar 

  45. Arriagada R, Bergman B, Dunant A, Le Chevalier T, Pignon JP, Vansteenkiste J. Cisplatin-based adjuvant chemotherapy in patients with completely resected non-small-cell lung cancer. N Engl J Med. 2004;350:351–60.

    Article  PubMed  Google Scholar 

  46. Winton T, Livingston R, Johnson D, Rigas J, Johnston M, Butts C, Cormier Y, Goss G, Inculet R, Vallieres E, et al. Vinorelbine plus cisplatin vs. observation in resected non-small-cell lung cancer. N Engl J Med. 2005;352:2589–97.

    Article  CAS  PubMed  Google Scholar 

  47. Douillard JY, Rosell R, De Lena M, Carpagnano F, Ramlau R, Gonzáles-Larriba JL, Grodzki T, Pereira JR, Le Groumellec A, Lorusso V, et al. Adjuvant vinorelbine plus cisplatin versus observation in patients with completely resected stage IB-IIIA non-small-cell lung cancer (Adjuvant Navelbine International Trialist Association [ANITA]): a randomised controlled trial. Lancet Oncol. 2006;7:719–27.

    Article  CAS  PubMed  Google Scholar 

  48. Altorki N, Wang X, Kozono D, Watt C, Landrenau R, Wigle D, Port J, Jones DR, Conti M, Ashrafi AS, et al. Lobar or sublobar resection for peripheral stage IA non-small-cell lung cancer. N Engl J Med. 2023;388:489–98.

    Article  PubMed  PubMed Central  Google Scholar 

  49. Zhu J, Fan Y, Xiong Y, Wang W, Chen J, Xia Y, Lei J, Gong L, Sun S, Jiang T. Delineating the dynamic evolution from preneoplasia to invasive lung adenocarcinoma by integrating single-cell RNA sequencing and spatial transcriptomics. Exp Mol Med. 2022;54:2060–76.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  50. Chen H, Carrot-Zhang J, Zhao Y, Hu H, Freeman SS, Yu S, Ha G, Taylor AM, Berger AC, Westlake L, et al. Genomic and immune profiling of pre-invasive lung adenocarcinoma. Nat Commun. 2019;10:5472.

    Article  PubMed  PubMed Central  Google Scholar 

  51. Hu X, Fujimoto J, Ying L, Fukuoka J, Ashizawa K, Sun W, Reuben A, Chow CW, McGranahan N, Chen R, et al. Multi-region exome sequencing reveals genomic evolution from preneoplasia to lung adenocarcinoma. Nat Commun. 2019;10:2978.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  52. Zhang C, Zhang J, Xu FP, Wang YG, Xie Z, Su J, Dong S, Nie Q, Shao Y, Zhou Q, et al. Genomic landscape and immune microenvironment features of preinvasive and early invasive lung adenocarcinoma. J Thorac Oncol. 2019;14:1912–23.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  53. Wang Z, Li Z, Zhou K, Wang C, Jiang L, Zhang L, Yang Y, Luo W, Qiao W, Wang G, et al. Deciphering cell lineage specification of human lung adenocarcinoma with single-cell RNA sequencing. Nat Commun. 2021;12:6500.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  54. Xing X, Yang F, Huang Q, Guo H, Li J, Qiu M, Bai F, Wang J. Decoding the multicellular ecosystem of lung adenocarcinoma manifested as pulmonary subsolid nodules by single-cell RNA sequencing. Sci Adv. 2021;7:eabd9738.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

Download references

Acknowledgements

We thank all the members of the Department of Thoracic Surgery, Tangdu Hospital, Fourth Military Medical University, Xi’an, China

Central Picture: The IA score can effectively stratify the risk of stage IA LUAD.

Funding

This study was funded by the National Natural Science Foundation of China (82002421), Health Care Special Research Project of Military Logistics Research Program (19BJZ20), Youth Talent Lifting Program of Shaanxi Association for Science and Technology (20230311), Key research and development plan of Shaanxi Province (2023-YBSF-318), Health Research Fund of Shaanxi Province (2021B004), Young Talent Program of Tangdu Hospital, New Clinical Technology Project of Tangdu Hospital (XJSXYW-2023041), Discipline Innovation Development Plan Project of Tangdu Hospital (2021LCYJ005), Partner Laboratory Cooperation Exchange Program of Air Force Medical University (2022HB010) and the "Open Competition Mechanism to Select the Best Candidates" Project of Tangdu Hospital (2022TDGS002).

Author information

Author notes

  1. Yanlu Xiong, Yongfu Ma and Kun Liu contributed equally to this article.

Authors and Affiliations

  1. Department of Thoracic Surgery, First Medical Center, Chinese PLA General Hospital and PLA Medical School, Beijing, China

    Yanlu Xiong, Yongfu Ma & Yang Liu

  2. Department of Thoracic Surgery, Tangdu Hospital, Fourth Military Medical University, Xi’an, China

    Yanlu Xiong, Jie Lei, Jinbo Zhao, Jianfei Zhu, Wenchen Wang, Miaomiao Wen, Xuejiao Wang, Ying Sun, Yabo Zhao, Yong Han & Tao Jiang

  3. Innovation Center for Advanced Medicine, Tangdu Hospital, Fourth Military Medical University, Xi’an, China

    Yanlu Xiong

  4. Department of Epidemiology, Ministry of Education Key Laboratory of Hazard Assessment and Control in Special Operational Environment, School of Public Health, Air Force Medical University, Xi’an, China

    Kun Liu

  5. Department of Thoracic Surgery, Shaanxi Provincial People’s Hospital, Xi’an, China

    Jianfei Zhu

  6. Department of Thoracic Surgery, Air Force Medical Center, Fourth Military Medical University, Beijing, China

    Yong Han

Authors
  1. Yanlu Xiong
    View author publications

    You can also search for this author in PubMed Google Scholar

  2. Yongfu Ma
    View author publications

    You can also search for this author in PubMed Google Scholar

  3. Kun Liu
    View author publications

    You can also search for this author in PubMed Google Scholar

  4. Jie Lei
    View author publications

    You can also search for this author in PubMed Google Scholar

  5. Jinbo Zhao
    View author publications

    You can also search for this author in PubMed Google Scholar

  6. Jianfei Zhu
    View author publications

    You can also search for this author in PubMed Google Scholar

  7. Wenchen Wang
    View author publications

    You can also search for this author in PubMed Google Scholar

  8. Miaomiao Wen
    View author publications

    You can also search for this author in PubMed Google Scholar

  9. Xuejiao Wang
    View author publications

    You can also search for this author in PubMed Google Scholar

  10. Ying Sun
    View author publications

    You can also search for this author in PubMed Google Scholar

  11. Yabo Zhao
    View author publications

    You can also search for this author in PubMed Google Scholar

  12. Yong Han
    View author publications

    You can also search for this author in PubMed Google Scholar

  13. Tao Jiang
    View author publications

    You can also search for this author in PubMed Google Scholar

  14. Yang Liu
    View author publications

    You can also search for this author in PubMed Google Scholar

Contributions

YLX, YL, TJ and YH had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis, including and especially any adverse effects. YLX, YFM, and KL contributed to the conception and analysis data. MMW, XJW, YS, and YBZ contributed to the acquisition and interpretation of data for the work. JL, JBZ, JFZ and WCW contributed to the design of the work. All authors contributed to the manuscript writing and revising, gave final approval of the manuscript as submitted and agreed to be accountable for all aspects of the work.

Corresponding authors

Correspondence to Yong Han, Tao Jiang or Yang Liu.

Ethics declarations

Ethics approval and consent to participate

For data obtained from public database (TCGA, GEO and SEER database), all ethics approvals and informed consents to participate were already obtained by the original studies. Besides, all procedures performed in studies involving human participants were conducted in accordance with the Declaration of Helsinki and its later amendments or comparable ethical standards, and approved by the Medical Ethics Committee of Tangdu Hospital, Fourth Military Medical University (K202107-19 and No. 202206-46). Informed consent was obtained from all subjects involved in the study.

Consent for publication

Informed consents were obtained from all individual participants included in the study.

Competing interests

The authors declare that they have no competing interests.

Additional information

Publisher's Note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Supplementary Information

Additional file 1: Table S1.

Details of the public datasets in this study.

Additional file 2: Table S2.

General information for fivefold cross-validation of stage IA in training set.

Additional file 3: Table S3.

Gene symbols and gene ids for 398 risk-related genes.

Additional file 4: Table S4.

Gene symbols and gene ids for 64 LASSO screened genes.

Additional file 5: Table S5.

Gene symbols and gene ids for 114 genes by LASSO within gene clusters.

Additional file 6: Table S6.

Risk factors for overall survival in stage IA lung adenocarcinoma.

Additional file 7: Table S7.

Risk factors for disease-free survival in stage IA lung adenocarcinoma.

Additional file 8: Figure S1.

Analysis of the effect of batch removal of 12 datasets. Usage of U-MAP to exhibit distribution of total samples (A) or LUAD (B) in 12 datasets calculated by gene expression profiles without Z-scaling for normalization; Usage of U-MAP to exhibit distribution of total samples (C) or LUAD (D) in 12 datasets calculated by gene expression profiles with Z-scaling for normalization; Usage of U-MAP to exhibit distribution of different histological sample in total 12 datasets (E) and specific single datasets (F, TCGA-LUAD; G, GSE31210, H, GSE68465; I, GSE30219; J, GSE42127) by gene expression profiles with Z-scaling for normalization. Usage of U-MAP to exhibit distribution of total samples in 12 datasets calculated by gene expression profiles by removeBatchEffect (K) and Combat (L) methods for batch effect removal. LUAD, lung adenocarcinoma; AD, adenocarcinoma; SQC, squamous cell carcinoma; Normal, normal tissue

Additional file 9: Figure S2.

Heatmaps for 398 risk-related genes and 64 LASSO screened genes. Unsupervised heatmaps for 398 risk-related genes (A) and 64 LASSO screened genes (B); Unsupervised heatmaps where smoking related residual values were removed in the gene expression matrix for 398 risk-related genes (C) and 64 LASSO screened genes (D). OR, Odds Ratio; LASSO, least absolute shrinkage and selection operator

Additional file 10: Figure S3.

Feature selection of genes representing different patterns. Consensus cumulative distribution function (CDF) (A) and Delta area (B) of consistent cluster analysis of genes through unsupervised clustering; consensus clustering of genes into four clusters (C); GO analysis of four clusters (D); prediction effect of 114 genes into the logistic regression model (classifying high-risk, H and low-risk, L) in training set (E) and external validation set (GSE30219) (F); prediction effect of 41 genes with significant coefficients into the logistic regression model (classifying high-risk, H and low-risk, L) in training set (G) and external validation set (GSE30219) (H); Heatmaps of 114 genes in the training set (original: I, smoking related residual values removed: K) and the validation set (J) respectively; GO analysis of 64 genes of IA score (L); Expression distribution of risk genes or oncogenes ((OR > 1) (M) and protective genes or tumor suppressor genes (OR < 1) (N) in early-stage LUAD (AIS, MIA and IAC). CDF, Consensus cumulative distribution function, OR, Odds Ratio; LASSO, least absolute shrinkage and selection operator; LUAD, lung adenocarcinoma; carcinoma in situ (AIS), microinvasive adenocarcinoma (MIA), invasive adenocarcinoma (IAC)

Additional file 11: Figure S4.

The effect of IA risk classification in the 8th TNM staging and the indication of IA score for chemotherapy sensitivity or not. (A) OS (left) and DSS (right) curves among IA1, IA2, IA3 and IB-II LUAD patients in SEER database; (B) OS (left) and DSS (right) curves between two categories of stage IA LUAD patients divided by receiving chemotherapy or not in SEER database (Some IA patients have no defined subclass); (C) OS (left) and DFS (right) curves between two categories of stage IA LUAD patients divided by receiving chemotherapy or not in GEO datasets possessing data about chemotherapy or not (GSE13213, GSE31210, GSE42127, GSE68465); (D) OS (left) and DSS (right) curves between two categories of stage IA3 LUAD patients divided by receiving chemotherapy or not in SEER database; (E) OS (left) and DFS (right) curves between two categories of stage IA L (low-risk type classified by IA score) LUAD patients divided by receiving chemotherapy or not in GEO datasets; (F) OS (left) and DFS (right) curves between two categories of stage IA H (high-risk type classified by IA score) LUAD patients divided by receiving chemotherapy or not in GEO datasets; (G) OS (left) and DSS (right) curves between two categories of stage IB-II LUAD patients divided by receiving chemotherapy or not in SEER database; (H) OS (left) and DFS (right) curves between two categories of stage IB-II LUAD patients divided by receiving chemotherapy or not in GEO datasets possessing data about chemotherapy or not; (I) OS (left) and DFS (right) curves between two categories of stage IB-II L (low-risk type classified by IA score) LUAD patients divided by receiving chemotherapy or not in GEO datasets; (J) OS (left) and DFS (right) curves between two categories of stage IB-II H (high-risk type classified by IA score) LUAD patients divided by receiving chemotherapy or not in GEO datasets. LUAD, lung adenocarcinoma; OS, overall survival; DFS, disease-free survival; DSS; disease-specific survival; SEER, Surveillance Epidemiology and End Results

Additional file 12.

Supplementary text.

Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data.

Reprints and permissions

About this article

Check for updates. Verify currency and authenticity via CrossMark

Cite this article

Xiong, Y., Ma, Y., Liu, K. et al. A gene-based score for the risk stratification of stage IA lung adenocarcinoma. Respir Res 25, 18 (2024). https://0-doi-org.brum.beds.ac.uk/10.1186/s12931-023-02647-4

Download citation

  • Received:

  • Accepted:

  • Published:

  • DOI: https://0-doi-org.brum.beds.ac.uk/10.1186/s12931-023-02647-4

Keywords

Respiratory Research

ISSN: 1465-993X

Contact us