即使在戒烟后，患者中发生肺腺癌（LUAD）的比例也在不断增加。在这里，我们的目的是确定吸烟是否会导致以前吸烟的患者的 LUAD 发生变化，这与不同的生物学和临床因素相对应。利用从从未吸烟的 LUAD 患者之间的差异表达基因开发的吸烟相关特征来训练随机森林模型 (RF)。来自 TCGA (n = 193) 和 BCCA (n = 69) 队列的吸烟者 (NS) 或当前吸烟者 (CS)。 RF 随后分别应用于 TCGA 和 MSKCC 队列中的 299 名和 131 名既往吸烟患者。 FS 被 RF 分类为 CS 样或 NS 样，并确定了与患者特征、生物学特征和临床结果的关联。我们阐明了 123 个基因特征，该特征在 RNA-seq 中对 NS 和 CS 进行了稳健分类（AUC = 0.85） ）和微阵列（AUC = 0.92）验证测试集。 RF 将 TCGA 队列中 213 名以前吸烟的患者分类为 CS 类患者，86 名患者分类为 NS 类患者。既往吸烟患者的 CS 样和 NS 样状态与患者特征相关性较差，但具有显着不同的生物学特征，包括肿瘤突变负荷、突变数量、诱变特征和免疫细胞群。 NS 样既往吸烟患者的总生存期分别比 TCGA 和 MSKCC 队列中的 CS 样患者长 17.5 个月和 18.6 个月。 既往吸烟且患有 LUAD 的患者具有异质性肿瘤生物学特征。这些患者可以根据吸烟诱导的基因表达进行分类，以了解预后和治疗选择的潜在生物学特征。© 2024。作者。

An increasing proportion of lung adenocarcinoma (LUAD) occurs in patients even after they have stopped smoking. Here, we aimed to determine whether tobacco smoking induced changes across LUADs from patients who formerly smoked correspond to different biological and clinical factors.Random forest models (RFs) were trained utilizing a smoking associated signature developed from differentially expressed genes between LUAD patients who had never smoked (NS) or currently smoked (CS) from TCGA (n = 193) and BCCA (n = 69) cohorts. The RFs were subsequently applied to 299 and 131 formerly smoking patients from TCGA and MSKCC cohorts, respectively. FS were RF-classified as either CS-like or NS-like and associations with patient characteristics, biological features, and clinical outcomes were determined.We elucidated a 123 gene signature that robustly classified NS and CS in both RNA-seq (AUC = 0.85) and microarray (AUC = 0.92) validation test sets. The RF classified 213 patients who had formerly smoked as CS-like and 86 as NS-like from the TCGA cohort. CS-like and NS-like status in formerly smoking patients correlated poorly with patient characteristics but had substantially different biological features including tumor mutational burden, number of mutations, mutagenic signatures and immune cell populations. NS-like formerly smoking patients had 17.5 months and 18.6 months longer overall survival than CS-like patients from the TCGA and MSKCC cohorts, respectively.Patients who had formerly smoked with LUAD harbor heterogeneous tumor biology. These patients can be divided by smoking induced gene expression to inform prognosis and underlying biological characteristics for treatment selection.© 2024. The Author(s).