Gene- and subtype-dependent prognostic impact of ras pathway mutations in acute myeloid leukemia: a cohort study of 2,500 patients Authors: Heng Shen, Yan Hui, Yuntao Liu, Shouyun Li, Dong Lin, Qiuyun Fang, Ying Wang, Benfa Gong, Chunlin Zhou, Kaiqi Liu, Guangji Zhang, Xiaoyuan Gong, Shaowei Qiu, Bingcheng Liu, Ying Wang, Yingchang Mi, Yan Li, Jianxiang Wang, Hui Wei Published: 2026 To the Editor: Acute myeloid leukemia (AML) is a hematologic malignancy driven by diverse molecular aberrations [1, 2]. Constitutive activation of signal transduction pathways is a hallmark of leukemogenesis [1, 3]. Among these, the RAS/MAPK signaling pathway is a central regulator of critical cellular processes, including proliferation, differentiation, and apoptosis [4]. RAS pathway mutations occur in approximately 25–30% of AML patients, but their prognostic significance remains controversial [5–7]. In AML, excessive activation of this pathway is predominantly driven by somatic mutations in RAS family members (NRAS and KRAS) or their upstream regulators and downstream effectors, such as PTPN11, CBL, and NF1 [8, 9]. Most prior studies were limited by relatively small sample sizes or the use of outdated risk stratification criteria [10–12]. Furthermore, the clinical impact of less common effectors like NF1 and CBL remains under-characterized. Therefore, we conducted a comprehensive analysis of NRAS, KRAS, PTPN11, CBL, and NF1 mutations in a large cohort of 2500 AML patients based on the latest European LeukemiaNet AML risk classification guidelines (ELN 2022). Between January 2015 and July 2025, a total of 2917 consecutive patients diagnosed with AML were screened at the Institute of Hematology and Blood Diseases Hospital of the Chinese Academy of Medical Sciences. After excluding patients aged > 65 years or those who did not receive intensive induction therapy, 2500 patients were included in the final analysis. The targeted next-generation sequencing (NGS) panels used at our center evolved over time: a 114-gene panel was used from 2015 to 2017, followed by a 175-gene panel (2018–2020), a 267-gene panel (2021–2023), and a 365-gene panel (2023–present). Detailed methods, treatment protocols, and statistical analyses are provided in the supplemental materials. The cohort (n = 2500) comprised 1,311 (52.4%) males and 1189 (47.6%) females, with a median age of 41.0 years (interquartile range [IQR], 31.0–51.0 years). A total of 899 (36.0%) patients harbored at least one RAS pathway mutation (RASmut). Patients were classified as wild-type (WT, n = 1601), co-mutation group (≥ 2 RAS pathway mutations, n = 253), NRAS-mutated (n = 400), KRAS-mutated (n = 106), PTPN11-mutated (n = 77), CBL-mutated (n = 35), or NF1-mutated (n = 28). Patients with NRAS, KRAS, and PTPN11 mutations had significantly higher white blood cell counts at diagnosis than WT patients, consistent with the proliferative role of RAS/MAPK signaling. ELN 2022 risk distribution differed by mutation: KRAS and NF1 mutations were enriched in adverse-risk AML, whereas NRAS mutations were more frequent in favorable-risk AML. Detailed baseline characteristics are presented in Table 1. Among RASmut patients, the most frequent alterations were FLT3 (27%), NPM1 (19%), KIT (14%), DNMT3A (13%), and WT1 (10%, Fig. S1A). Co-mutation analysis demonstrated that NRAS mutations were mutually exclusive with other RAS pathway mutations (all p < 0.001, Fig. S1B), suggesting functional redundancy within the RAS pathway. PTPN11 mutations were associated with NPM1 and DNMT3A mutations, whereas KRAS mutations were mutually exclusive with NPM1. Co-mutation analysis results are shown in Fig. S1B. Table 1Baseline characteristics of AML patients with RAS pathway gene mutations.CharacteristicsTotal cohort (N = 2500)Wild-type (n = 1601)Co-mutation (n = 253)NRASmut (n = 400)KRASmut (n = 106)PTPN11mut (n = 77)CBLmut (n = 35)NF1mut (n = 28)p valueSex (n, %)0.327Male1311 (52.4)833 (52.0)136 (53.8)220 (55.0)55 (51.9)32 (41.6)22 (62.9)13 (46.4)Female1189 (47.6)768 (48.0)117 (46.2)180 (45.0)51 (48.1)45 (58.4)13 (37.1)15 (53.6)Age at diagnosis (median, IQR)41.0 (31.0, 51.0)42.0 (31.0, 51.0)41.0 (31.0, 50.0)38.5 (28.0, 49.0)37.5 (27.3, 46.5)43.0 (32.0, 53.0)43.0 (28.0, 51.0)39.0 (31.8, 52.5)0.038WBC, ×109/L (median, IQR)14.4 (4.3, 42.4)12.2 (3.71, 38.6)17.9 (8.00, 44.8)22.3 (6.4, 49.9)19.9 (5.5, 45.5)18.4 (4.6, 55.7)9.9 (3.0, 30.5)8.8 (4.0, 19.9)< 0.001Hb, g/L (median, IQR)83.0 (70.0, 99.0)84.0 (71.0, 100.0)80.0 (69.0, 95.0)80.0 (68.8, 99.4)87.0 (70.0, 101.5)80.0 (69.0, 100.0)81.0 (71.0, 95.0)74.0 (68.8, 100.5)0.187Plt, ×109/L (median, IQR)41.0 (25.0, 74.0)40.0 (25.0, 75.0)40.0 (24.0, 76.0)42.0 (23.8, 66.0)41.0 (27.0, 68.0)68.0 (40.0, 96.0)40.0 (20.5, 58.0)40.0 (24.3, 84.8)0.001Mutation (n, %)FLT3681 (27.2)435 (27.2)76 (30.0)102 (25.5)27 (25.5)29 (37.7)3 (8.6)9 (32.1)0.052KIT335 (13.4)211 (13.2)38 (15.0)60 (15.0)12 (11.3)5 (6.5)5 (14.3)4 (14.3)0.528DNMT3A313 (12.5)197 (12.3)40 (15.8)41 (10.2)7 (6.6)20 (26.0)4 (11.4)4 (14.3)0.002WT1225 (9.0)135 (8.4)18 (7.1)42 (10.5)12 (11.3)10 (13.0)6 (17.1)2 (7.1)0.240IDH2178 (7.1)129 (8.1)15 (5.9)27 (6.8)2 (1.9)3 (3.9)1 (2.9)1 (3.6)0.131TET2161 (6.4)125 (7.8)10 (4.0)17 (4.2)2 (1.9)3 (3.9)3 (8.6)1 (3.6)0.014IDH1125 (5.0)73 (4.6)12 (4.7)27 (6.8)1 (0.9)11 (14.3)0 (0.0)1 (3.6)0.001ASXL1118 (4.7)69 (4.3)11 (4.3)19 (4.8)7 (6.6)4 (5.2)3 (8.6)5 (17.9)0.036GATA2112 (4.5)71 (4.4)13 (5.1)22 (5.5)2 (1.9)4 (5.2)0 (0.0)0 (0.0)0.429RUNX1103 (4.1)51 (3.2)17 (6.7)17 (4.2)4 (3.8)8 (10.4)3 (8.6)3 (10.7)0.002U2AF191 (3.6)44 (2.7)19 (7.5)12 (3.0)8 (7.5)3 (3.9)3 (8.6)2 (7.1)0.001CSF3R86 (3.4)55 (3.4)5 (2.0)19 (4.8)3 (2.8)1 (1.3)2 (5.7)1 (3.6)0.490WHO 2022(n, %)< 0.001AML with CEBPAbzip-inf mutation327 (15.9)256 (19.9)15 (6.6)49 (14.8)3 (3.2)2 (3.1)0 (0.0)2 (7.4)AML with RUNX1::RUNX1T1 fusion438 (21.3)324 (25.2)23 (10.2)64 (19.3)9 (9.5)2 (3.1)14 (46.7)2 (7.4)AML with NPM1 mutation469 (22.8)306 (23.8)53 (23.5)58 (17.5)7 (7.4)36 (55.4)3 (10.0)6 (22.2)AML with KMT2A rearrangement192 (9.3)87 (6.8)34 (15.0)30 (9.1)29 (30.5)9 (13.8)3 (10.0)0 (0.0)AML with CBFB::MYH11 fusion204 (9.9)60 (4.7)51 (22.6)67 (20.2)21 (22.1)0 (0.0)1 (3.3)4 (14.8)AML with NUP98 rearrangement39 (1.9)28 (2.2)4 (1.8)3 (0.9)2 (2.1)0 (0.0)1 (3.3)1 (3.7)AML with DEK::NUP214 fusion14 (0.7)10 (0.8)1 (0.4)1 (0.3)2 (2.1)0 (0.0)0 (0.0)0 (0.0)AML with MECOM rearrangement17 (0.8)5 (0.4)4 (1.8)3 (0.9)3 (3.2)2 (3.1)0 (0.0)0 (0.0)AML with BCR::ABL1 fusion10 (0.5)8 (0.6)1 (0.4)0 (0.0)0 (0.0)0 (0.0)0 (0.0)1 (3.7)AML with myelodysplasia-related changes349 (16.9)201 (15.6)40 (17.7)56 (16.9)19 (20.0)14 (21.5)8 (26.7)11 (40.7)2022 ELN risk stratification (n, %)< 0.001Favorable1221 (48.8)786 (49.1)131 (51.8)211 (52.8)37 (34.9)27 (35.1)17 (48.6)12 (42.9)Intermediate728 (29.1)520 (32.5)52 (20.6)103 (25.8)18 (17.0)26 (33.8)7 (20.0)2 (7.1)Adverse551 (22.0)295 (18.4)70 (27.7)86 (21.5)51 (48.1)24 (31.2)11 (31.4)14 (50.0) Baseline characteristics of AML patients with RAS pathway gene mutations. In our cohort, 44 (1.8%) patients harbored RAS pathway mutations without recurrent co-mutations. No significant differences in OS or EFS were observed between patients with isolated and non-isolated RAS pathway mutations (5-year OS: 55.3% [95% CI, 40.8–75.1%] vs. 63.4% [95% CI, 59.4–67.8%], p = 0.400; 5-year EFS: 53.8% [95% CI, 38.5–75.3%] vs. 51.1% [95% CI, 46.6–56.0%], p = 0.950, Fig. S2A, B). Notably, all 44 patients with isolated RAS pathway mutations were classified as intermediate-risk according to the 2022 ELN guidelines. There were no significant differences in OS or EFS between patients with isolated and non-isolated RAS pathway mutations (5-year OS: 55.3% [95% CI, 40.8–75.1%] vs. 51.9% [95% CI, 42.9–62.9%], p = 0.690; 5-year EFS: 53.8% [95% CI, 38.5–75.3%] vs. 38.4% [95% CI, 29.2–50.5%], p = 0.310, Fig. S2C, D). Then RAS pathway mutations were considered as a whole. There was no significant difference in OS or EFS between the RASmut and WT group (5-year OS: 62.9% [95% CI, 58.9–67.1%] vs. 58.8% [95% CI, 55.6–62.1%], p = 0.670; 5-year EFS: 51.4% [95% CI, 47.1–56.2%] vs. 48.5% [95% CI, 45.5–51.8%], p = 0.510; Fig. 1A, B). When patients were stratified by the 2022 ELN guidelines, RAS pathway mutations were associated with superior OS (5-year OS: 79.0% [95% CI, 74.2–84.1%] vs. 69.7% [95% CI, 65.5–74.2%], p = 0.036; Fig. S3A) in the favorable-risk subgroup, whereas no significant difference was observed in EFS (5-year EFS: 64.1% [95% CI, 58.1–70.8%] vs. 59.3% [95% CI, 55.1–63.9%], p = 0.120; Fig. S3B). No significant associations with either OS or EFS were observed in the intermediate- or adverse-risk subgroups (Fig. S3C–F). To identify which molecular subtype drove this effect, we performed subtype analyses. We found that the superior OS and EFS associated with RAS mutations were restricted to patients harboring the RUNX1::RUNX1T1 fusion (5-year OS: 84.0% [95% CI, 75.8–93.1%] vs. 66.0% [95% CI, 59.6–73.1%], p = 0.016; 5-year EFS: 67.4% [95% CI, 55.8–81.4%] vs. 57.0% [95% CI, 50.8–64.0%], p = 0.039; Fig. S4A, B). No such effect was observed in CBFB::MYH11, NPM1-mutated, or CEBPA bZIP in-frame mutated (CEBPAbZIP-inf) AML (Fig. S4C–H). Kaplan-Meier survival analyses according to RAS pathway mutations in AML. A, B OS (A) and EFS (B) in the overall cohort stratified by RAS pathway mutation status. C–H OS and EFS stratified by individual RAS pathway gene mutations: NRAS (C, D), PTPN11 (E, F), and NF1 (G, H). We further evaluated the prognostic impact of individual RAS pathway mutations. NRAS mutations were associated with favorable outcomes, showing significantly superior OS and EFS (5-year OS: 67.3% [95% CI, 62.7–72.2%] vs. 57.8% [95% CI, 54.9–60.9%], p = 0.003; 5-year EFS: 53.9% [95% CI, 48.7–59.7%] vs. 48.1% [95% CI, 45.2–51.1%], p = 0.010, Fig. 1C, D). PTPN11 mutations were associated with inferior EFS and showed a trend toward worse OS (5-year OS: 54.6% [95% CI, 45.9–65.0%] vs. 60.5% [95% CI, 57.9–63.2%], p = 0.062; 5-year EFS: 45.3% [95% CI, 36.9–55.6%] vs. 49.9% [95% CI, 47.3–52.6%], p = 0.017; Fig. 1E, F). Similarly, NF1 mutations showed a trend toward inferior OS (5-year OS: 60.5% [95% CI, 45.0–80.2%] vs. 60.2% [95% CI, 57.6–62.8%], p = 0.090; Fig. 1G), while no significant difference was observed in EFS (Fig. 1H). Neither KRAS nor CBL mutations showed a significant association with outcomes. Since NRAS, PTPN11, and NF1 mutations demonstrated prognostic relevance in our cohort, we further evaluated their specific roles within the ELN 2022 risk subgroups. In the favorable-risk group, NRAS mutations were associated with significantly superior OS and EFS (5-year OS: 80.5% [95% CI, 75.2–86.1%] vs. 70.2% [95% CI, 66.3–74.4%], p = 0.018, Fig. S5A; 5-year EFS: 65.0% [95% CI, 58.0–72.8%] vs. 59.6% [95% CI, 55.6–63.9%], p = 0.034, Fig. S5B), whereas no survival differences were observed in the intermediate- or adverse-risk groups. Subtype analysis showed that NRAS mutations were associated with improved OS and EFS in RUNX1::RUNX1T1 AML (5-year OS: 85.5% [95% CI, 76.4–95.7%] vs. 66.8% [95% CI, 60.6–73.6%], p = 0.008, Fig. S6A; 5-year EFS: 70.4% [95% CI, 57.2–86.7%] vs. 57.1% [95% CI, 51.1–63.8%], p = 0.010, Fig. S6B), and superior OS in NPM1-mutated AML (5-year OS: 70.8% [95% CI, 59.7–83.8%] vs. 59.3% [95% CI, 53.2–66.2%], p = 0.041, Fig. S6C). PTPN11 mutations showed a tendency toward shorter EFS in the favorable-risk group (5-year EFS: 53.5% [95% CI, 40.7–70.3%] vs. 61.5% [95% CI, 57.9–65.3%], p = 0.066, Fig. S5D), without significant differences in OS (Fig. S5C), and no prognostic impact in the intermediate- or adverse-risk groups. Subtype analysis confirmed that PTPN11 mutations were not associated with OS and EFS in RUNX1::RUNX1T1, CBFB::MYH11, or NPM1-mutated AML. NF1 mutations were associated with significantly poorer OS in the adverse-risk group (5-year OS: 30.1% [95% CI, 11.2–80.6%] vs. 41.1% [95% CI, 35.4–47.7%], p = 0.013, Fig. S5E), with no impact on EFS (Fig. S5F). No significant prognostic effect was observed in the favorable- or intermediate-risk groups. Further subtype analysis showed no significant prognostic impact of NF1 mutations in AML with myelodysplasia-related changes (AML-MRC), NPM1-mutated, or CBFB::MYH11 AML. We subsequently performed multivariate analysis incorporating consolidation regimen, WBC, Hb, PLT, NRAS mutation, PTPN11 mutation, transplantation, and ELN 2022 risk classification, since previous work has suggested that AML patients with RAS mutations may benefit from higher-dose cytarabine than those without RAS mutations [5]. The results showed that PTPN11 mutations lost their prognostic significance for both OS (HR, 0.924 [95% CI, 0.696–1.227]; p = 0.586) and EFS (HR, 1.212 [95% CI, 0.956–1.538]; p = 0.112, Table S1). NRAS mutations remained independently associated with superior OS (HR, 0.728 [95% CI, 0.600–0.883]; p = 0.001), but not EFS (HR, 0.873 [95% CI, 0.744–1.024]; p = 0.095, Table S1). Despite the strengths of this study, including the massive sample size and broad gene panel, several limitations should be noted. First, as a retrospective analysis, heterogeneity in therapies may have influenced long-term outcomes. Second, while we expanded our focus to include CBL and NF1, other rare RAS-pathway effectors such as BRAF or SOS1 were not the primary focus, which may slightly underestimate the total pathway burden. Third, post-treatment follow-up samples were not available in our cohort, so assessment of measurable residual disease (MRD) and analysis of variant allele frequency (VAF) dynamics in RAS pathway mutations were impossible. In summary, our findings highlight the prognostic heterogeneity of RAS pathway mutations in AML, as genes within this pathway, including NRAS, KRAS, PTPN11, NF1, and CBL, have distinct clinical and prognostic implications, likely in part reflecting differences in their co-mutational backgrounds. NRAS mutation identifies a subgroup with favorable outcomes, particularly in RUNX1::RUNX1T1 AML. These findings suggest that future risk-stratification models and clinical trials should incorporate gene-specific and genomic context-dependent effects rather than relying on a binary “RAS-mutated” classification. Supplementary information Supplementary Methods Supplementary Table 1 Supplementary figure legend Supplementary Figure 1 Supplementary figure 2 Supplementary figure 3 Supplementary figure 4 Supplementary figure 5 Supplementary figure 6 Author contributions H.S. and Y.H. contributed equally to this work. H.W., J.X.W., and Y.L. conceptualized the study. H.S. and Y.H. analyzed the data and wrote the manuscript. Y.T.L., S.Y.L., D.L., Q.Y.F., Y.W. (7th author), B.F.G., C.L.Z., K.Q.L., G.J.Z., X.Y.G., S.W.Q., B.C.L., Y.W. (15th author) and Y.C.M. collected the clinical information of the patients. Funding National Key Research and Development Program of China (2023YFC2508900), National Natural Science Foundation of China (82370183), CAMS Innovation Fund for Medical Sciences (2025-I2M-KJ-026), Tian Jin Natural Science Foundation (23JCZXJC00310), Tianjin Municipal Science and Technology Commission Grant (25ZXRKSY00050), Beijing Xisike Clinical Oncology Research Foundation (Y-SYBLD2022ZD-0031). Data availability The data that support the findings of this study are available from the corresponding author upon reasonable request. Competing interests The authors declare no competing interests. Ethics approval and consent to participate All methods were performed in accordance with the relevant guidelines and regulations. The study was approved by the Blood Diseases Hospital Ethics Committee and conducted in accordance with the Declaration of Helsinki. Written informed consent for treatment was obtained from all patients. Footnotes Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. These authors contributed equally: Heng Shen, Yan Hui. Contributor Information Yan Li, Email: liyan1@ihcams.ac.cn. Jianxiang Wang, Email: wangjx@ihcams.ac.cn. Hui Wei, Email: weihui@ihcams.ac.cn. Supplementary information The online version contains supplementary material available at 10.1038/s41408-026-01532-9. References Associated Data Supplementary Materials Supplementary Methods Supplementary Table 1 Supplementary figure legend Supplementary Figure 1 Supplementary figure 2 Supplementary figure 3 Supplementary figure 4 Supplementary figure 5 Supplementary figure 6 Data Availability Statement The data that support the findings of this study are available from the corresponding author upon reasonable request.