A genotype-guided prediction model for the incidence of persistent acute kidney injury following lung transplantation Authors: Wenwen Du, Xiaoxing Wang, Dan Zhang, Wenqian Chen, Xianbo Zuo, Pengmei Li Published: 2024 ABSTRACT AbstractBackgroundThis study aimed to develop a nomogram for predicting persistent renal dysfunction in acute kidney injury (AKI) following lung transplantation (LTx).MethodA total of 229 LTx patients were enrolled, and genotyping for 153 single nucleotide polymorphisms (SNPs) was performed. The cohort was randomly divided into training (n = 183) and validation (n = 46) sets in an 8:2 ratio. Statistically significant SNPs identified through pharmacogenomic analysis were combined with clinical factors to construct a comprehensive prediction model for persistent AKI using multivariate logistic regression analysis. Discrimination and calibration analyses were conducted to evaluate the performance of the model. Decision curve analysis was used to assess its clinical utility. Due to the small sample size, bootstrap internal sampling with 500 iterations was adopted for validation to prevent overfitting of the model.ResultsThe final nomogram comprised nine predictors, including body mass index, thrombin time, tacrolimus initial concentration, rs757210, rs1799884, rs6887695, rs1494558, rs2069762 and rs2275913. In the training set, the area under the receiver operating characteristic curve of the nomogram was 0.781 (95%CI: 0.715–0.846), while in the validation set it was 0.698 (95%CI: 0.542–0.855), indicating good model fit. As demonstrated by 500 Bootstrap internal sampling validations, the model has high discrimination and calibration. Additionally, decision curve analysis confirmed its clinical applicability.ConclusionThis study presents a genotype-guided nomogram that can be used to assess the risk of persistent AKI following LTx and may assist in guiding personalized prevention strategies in clinical practice.Supplementary InformationThe online version contains supplementary material available at 10.1186/s12882-024-03871-w.Keywords: Persistent acute kidney injury, Lung transplantation, Tacrolimus, Prediction model Introduction Acute kidney injury (AKI) is a common complication following lung transplantation (LTx), affecting 36.6% to 84.9% of LTx patients and exacerbating their short- and long-term outcomes [1–5]. Persistent AKI, defined as an AKI that lasts for more than 48 h from its onset, has been consistently associated with increased morbidity and mortality rates post-LTx [5–7]. The timely recognition of persistent AKI is crucial to facilitate a more comprehensive diagnostic and management approach, thereby potentially mitigating further renal impairment and reducing the associated risk of mortality [6]. The development of persistent AKI after LTx is influenced by various factors including pre-operative body mass index (BMI), pulmonary hypertension, intraoperative or post-operative severe hypotension, post-operative multiple organ dysfunction, mechanical ventilation and nephrotoxic agents [5, 8]. However, no study has yet combined these potential risk factors to construct a prediction model for the occurrence of persistent AKI after LTx; therefore, it is meaningful to investigate this aspect further. Additionally, genetic factors have been suggested as potentially responsible for AKI susceptibility and severity, however, there are currently no reports on the correlation between genetic factors and persistent AKI after LTx [9, 10]. Gaining a better understanding of the genetics underlying predisposition to persistent AKI may provide valuable insights into novel approaches for preventing this condition. The objective of this study is to identify single nucleotide polymorphisms (SNPs) associated with an increased susceptibility to persistent AKI following LTx, aiming to develop a genotype-guided predictive model that integrates these genetic variants with significant clinical risk factors, thereby enhancing the accuracy of post-LTx persistent AKI risk assessment. By utilizing our predictive nomogram, healthcare providers can more precisely identify patients at elevated risk for persistent AKI, enabling timely and targeted clinical intervention. Methods Study design and population In this single center, retrospective study, patients who underwent LTx in China-Japan Friendship Hospital from August 2016 to April 2023 were enrolled according to predefined criteria. The inclusion criteria comprised of: (1) first-time LTx recipients; (2) age ≥ 18; and (3) maintenance immunosuppression with tacrolimus (Tac). Exclusion criteria were: (1) unavailability of blood samples for genotyping; (2) received continuous renal replacement therapy (CRRT) before LTx; (3) survival less than 14 days post-transplantation; and (4) missing data. Immunosuppressive regimens Patients were administered methylprednisolone as a monotherapy or in combination with basiliximab for induction therapy. Methylprednisolone was administered intraoperatively at a dosage range of 500 to 1000 mg. Basiliximab was administered at a dosage of 20 mg, both preoperatively and on post-operative day (POD) 4. The maintenance immunosuppressive regimen comprised Tac, mycophenolate mofetil, and prednisone. Tac was initiated on POD1 at a starting dose of 2 mg twice daily. Subsequent dose adjustments were guided by trough concentration measurements. The initial mycophenolate mofetil dosage was set at 500 mg twice daily. Subsequent dose adjustments were made in response to adverse drug reactions. Prednisone was administered intravenously at a dosage of 1 mg/(kg·d) from POD 1 to 3, followed by a transition to oral administration at a reduced dosage of 0.5 mg/(kg·d). The prednisone dosage was then tapered to achieve a long-term maintenance dose within the range of 5–10 mg daily. Genotyping Residual blood samples, originally collected for the routine assessment of Tac levels after LTx, were repurposed for genotype analysis. A detailed list of the SNPs examined in this study is provided in Supplementary Table S1. For genotyping, a 0.5 mL blood sample was collected and genomic DNA was extracted from EDTA-anticoagulated peripheral blood using a commercially available DNA purification kit (EasyPure Blood Genomic DNA Kit, Transgene Biotech, Beijing, China). Samples with a call rate > 95% were genotyped using an ASACHIA Bead chip (Infinium Asian Screening Array-China Health Industry Alliance) according to the Illumina Infinium HTS chip standard operating procedures. AKI definition AKI was determined according to the ‘Kidney Disease: Improving Global Outcomes’ (KDIGO) Clinical Practice Guideline, basing on serum creatinine (SCr) obtained from the transplantation date until 7 days after transplantation. According to the guideline, AKI was defined as any of the followings: (1) increase in SCr by ≥ 26.5 μmol/L within 48 h; (2) increase in SCr to ≥ 1.5 times baseline in the first 7 post-operative days. Baseline SCr was defined as the SCr value obtained prior to LTx. Persistent AKI was defined as incomplete recovery from AKI (SCr above KDIGO criteria) in the first 7 days. Data collection for persistent AKI analysis In this study, a thorough examination of patients’ electronic medical records was performed to gather data pertinent to persistent AKI. The data collected included a wide range of parameters: baseline demographics, pre-operative, intra-operative, and post-operative information. Baseline demographics The demographic data collected at baseline encompassed age, sex, height, and weight. Pre-operative data Pre-operative data included the indication for LTx, comorbidities, the American Society of Anesthesiologists (ASA) score, and baseline laboratory parameters. These laboratory parameters comprised alanine aminotransferase (ALT), aspartate aminotransferase (AST), total protein (TP), albumin (ALB), total bilirubin (Tbil), direct bilirubin (Dbil), serum creatinine (SCr), uric acid (UA), urea, glucose, white blood cell (WBC) count, neutrophils count, lymphocyte count, red blood cell (RBC) count, hemoglobin, hematocrit, potassium, sodium, calcium, chlorine, prothrombin time (PT), prothrombin activity (PTA), international normalized ratio (INR), fibrinogen (Fib), activated partial thromboplastin time (APTT), thrombin time (TT), D-D dimer (DD), fibrin degradation products (FDP). Intra-operative data Intra-operative data included the type of transplantation, the use of basiliximab and extracorporeal membrane oxygenation (ECMO), administration of colloid, crystalloid, blood products, and autologous blood, as well as the volume of urine and blood loss, and the dosage of furosemide. Post-operative data Post-operative data focused on the daily average dose in the first post-operative week and the initial concentration of Tac, and 90-day survival status. Statistical analysis Data was processed using R (version 4.3.1) software. The assessment of linkage disequilibrium among SNPs was performed using PLINK software. Distribution of data was assessed using Shapiro–Wilk test. Continuous variables were expressed as mean ± standard deviation (SD) for normally distributed continuous variables or median (interquartile range, IQR) for non-normally distributed continuous variables. Categorical variables were presented as count (percentage). For laboratory test results, variables were converted into categorical variables using the laboratory cut-off values. The association analysis between candidate SNPs and the occurrence of persistent AKI was conducted using three genetic models: codominant, dominant, and recessive. Univariate logistic regression analysis was employed to identify potential clinical risk factors for persistent AKI. Clinical parameters with a p-value < 0.1, along with candidate SNPs that demonstrated a significant association with persistent AKI, were considered for further analysis. Multivariate logistic regression analysis was utilized to determine the final predictors and establish the prediction nomogram from the data of the training set. The predictive accuracy of the nomogram was evaluated using the area under the curve (AUC) of the receiver operating characteristic (ROC) analysis. In addition, we used the F1 score for performance measures. The variance between the predicted value derived from the nomogram and the actual value was assessed using a calibration curve and the Hosmer–Lemeshow goodness of fit test (HL test). Furthermore, the clinical utility of the predictive nomogram was examined through decision curve analysis (DCA). The rms package in R software was used to develop a nomogram, which visually represents the prediction model based on the multivariate logistic regression analysis. To prevent overfitting, bootstrap resampling was performed 500 times to evaluate the internal validation of predictive efficiency. The significance level was set at p = 0.05. Results Demographics A total of 240 patients underwent LTx and participated in genotyping between August 2016 to April 2023. From this study population, two patients were excluded for being under the age of 18, two patients were excluded for using cyclosporine, five patients were excluded due to missing data, one patient was excluded for survived less than 14 days and one patient was excluded for receiving CRRT before LTx. Finally, a total of 229 patients were included in the analysis and were randomly divided into two subsets: a training set (n = 183, 80%) and an internal validation set (n = 46, 20%). The baseline characteristics are presented in Table 1. In the training and validation sets, persistent AKI was observed in Eighty-two (44.81%) and eighteen (39.13%) patients respectively. Table 1Basic characteristics of study population in the training and validation setsParametersAll, n = 229Training set, n = 183Validation set, n = 46p-valueAge0.790 < 60108 (47.16%)85 (46.45%)23 (50.00%) > 60121 (52.84%)98 (53.55%)23 (50.00%)Sex0.432 Male186 (81.22%)151 (82.51%)35 (76.09%) Female43 (18.78%)32 (17.49%)11 (23.91%)BMI0.130 < 18.5 kg/m247 (20.52%)35 (19.13%)12 (26.09%) 18.5–23.9 kg/m2105 (45.85%)90 (49.18%)15 (32.61%) > 23.9 kg/m277 (33.62%)58 (31.69%)19 (41.30%)Indication for LTx0.112 ILD178 (77.73%)144 (78.69%)34 (73.91%) COPD20 (8.73%)15 (8.20%)5 (10.87%) Bronchiectasis10 (4.37%)10 (5.46%)0 (0.00%) BO10 (4.37%)8 (4.37%)2 (4.35%) Others11 (4.80%)6 (3.28%)5 (10.87%)Pre-operative comorbidities Hypertension44 (19.21%)37 (20.22%)7 (15.22%)0.575 Hyperlipidemia17 (7.42%)12 (6.56%)5 (10.87%)0.346 T2DM64 (27.95%)50 (27.32%)14 (30.43%)0.813 Cardiovascular disease40 (17.47%)31 (16.94%)9 (19.57%)0.840Pre-operative parameters ALT > 40 IU/L36 (15.72%)27 (14.75%)9 (19.57%)0.565 AST > 42 IU/L22 (9.61%)17 (9.29%)5 (10.87%)0.780TP, g/L0.941 < 6020 (8.73%)17 (9.29%)3 (6.52%) 60–80194 (84.72%)154 (84.15%)40 (86.96%) > 8015 (6.55%)12 (6.56%)3 (6.52%) ALB < 35 g/L30 (13.10%)28 (15.30%)2 (4.35%)0.085Tbil, μmol/L0.338 < 541 (17.90%)30 (16.39%)11 (23.91%) 5–21183 (79.91%)148 (80.87%)35 (76.09%) > 215 (2.18%)5 (2.73%)0 (0.00%) Dbil > 7 μmol/L6 (2.62%)6 (3.28%)0 (0.00%)0.603 Glu > 6.11 mmol/L143 (62.45%)110 (60.11%)33 (71.74%)0.199Urea, mmol/L0.437 < 2.788 (3.49%)7 (3.83%)1 (2.17%) 2.78–7.85187 (81.66%)146 (79.78%)41 (89.13%) > 7.8534 (14.85%)30 (16.39%)4 (8.70%) SCr > 106 μmol/L1 (0.44%)1 (0.55%)0 (0.00%)1.000UA, μmol/L0.769 < 15034 (14.85%)29 (15.85%)5 (10.87%) 150–420176 (76.86%)139 (75.96%)37 (80.43%) > 42019 (8.30%)15 (8.20%)4 (8.70%)Potassium, mmol/L0.392 < 3.511 (4.80%)7 (3.83%)4 (8.70%) 3.5–5.5217 (94.76%)175 (95.63%)42 (91.30%) > 5.51 (0.44%)1 (0.55%)0 (0.00%)Sodium, mmol/L0.580 < 13525 (10.92%)18 (9.84%)7 (15.22%) 135–145197 (86.03%)159 (86.89%)38 (82.61%) > 1457 (3.06%)6 (3.28%)1 (2.17%)Chlorine, mmol/L0.297 < 902 (0.87%)1 (0.55%)1 (2.17%) 90–110221 (96.51%)176 (96.17%)45 (97.83%) > 1106 (2.62%)6 (3.28%)0 (0.00%)Calcium, mmol/L0.110 < 217 (7.42%)11 (6.01%)6 (13.04%) 2–2.75210 (91.70%)171 (93.44%)39 (84.78%) > 2.752 (0.87%)1 (0.55%)1 (2.17%) WBC > 9.5 × 109 cells/L86 (37.55%)66 (36.07%)20 (43.48%)0.449 NEUT > 6.3 × 109 cells/L100 (43.67%)80 (43.72%)20 (43.48%)1.000Lympho, × 109 cells/L0.623 < 1.154 (23.58%)41 (22.40%)13 (28.26%) 1.1–3.2166 (72.49%)135 (73.77%)31 (67.39%) > 3.29 (3.93%)7 (3.83%)2 (4.35%)RBC, × 1012 cells/L0.470 < 4.3117 (51.09%)94 (51.37%)23 (50.00%) 4.3–5.8107 (46.72%)86 (46.99%)21 (45.65%) > 5.85 (2.18%)3 (1.64%)2 (4.35%)Hemoglobin, g/L0.388 < 130109 (47.60%)90 (49.18%)19 (41.30%) 130–175117 (51.09%)91 (49.73%)26 (56.52%) > 1753 (1.31%)2 (1.09%)1 (2.17%)Hematocrit, %0.293 < 40123 (53.71%)100 (54.64%)23 (50.00%) 40–5099 (43.23%)79 (43.17%)20 (43.48%) > 507 (3.06%)4 (2.19%)3 (6.52%)INR1.000 < 0.854 (1.75%)3 (1.64%)1 (2.17%) 0.85–1.5224 (97.82%)179 (97.81%)45 (97.83%) > 1.51 (0.44%)1 (0.55%)0 (0.00%)Fib, g/L0.437 < 25 (2.18%)5 (2.73%)0 (0.00%) 2–492 (40.17%)76 (41.53%)16 (34.78%) > 4132 (57.64%)102 (55.74%)30 (65.22%)APTT, s0.050 < 283 (1.31%)2 (1.09%)1 (2.17%) 28–43.5187 (81.66%)145 (79.23%)42 (91.30%) > 43.539 (17.03%)36 (19.67%)3 (6.52%) TT > 21 s9 (3.93%)8 (4.37%)1 (2.17%)0.691 FDP > 5 μg/mL37 (16.16%)31 (16.94%)6 (13.04%)0.676Intra-operative parameters Transplant type1.000 Unilateral88 (38.43%)70 (38.25%)18 (39.13%) Bilateral141 (61.57%)113 (61.75%)28 (60.87%) Operation time, min330 (235–400)330 (228–404)330 (241–384)0.983ASA score0.450 23 (1.31%)2 (1.09%)1 (2.17%) 398 (42.79%)79 (43.17%)19 (41.30%) 4126 (55.02%)101 (55.19%)25 (54.35%) 52 (0.87%)1 (0.55%)1 (2.17%)ECMO support176 (76.86%)139 (75.96%)37 (80.43%)0.654Furosemide dosage, mg0 (0–10)0 (0–10)0 (0–0)0.158Blood output, mL400 (300–800)400 (300–800)400 (300–575)0.420Urine output, mL660 (400–1200)660 (400–1225)660 (418–1108)0.929Total output volume, mL1250 (800–1950)1260 (800–1980)1200 (755–1782)0.444Blood product transfusion, mL0 (0–800)0 (0–800)0 (0–550)0.710Autologous blood transfusion, mL50 (0–100)50 (0–100)50 (0–100)0.990Crystalloid administration, mL2600 (2000–3450)2500 (1900–3425)2900 (2000–3475)0.399Colloid administration, mL250 (100–500)250 (100–500)250 (50–400)0.173Total input volume, mL3150 (2350–4500)3150 (2350–4500)3100 (2488–4488)0.966Basiliximab use127 (55.46%)98 (53.55%)29 (63.04%)0.321Prednisone dose750 (500–750)750 (500–750)750 (500–750)0.745Tac data Daily average Tac dose2.29 (1.57–3.07)2.14 (1.57–3.14)2.39 (1.50–2.93)0.937 Initial concentration > 6.5 ng/mL65 (28.38%)56 (30.60%)9 (19.57%)0.193Abbreviations AKI Acute kidney injury, ALB Albumin, ALT Alanine aminotransferase, APTT Activated partial thromboplastin time, ASA American Society of Anesthesiologists, AST Aspartate aminotransferase, BMI Body mass index, BO Bronchiolitis obliterans, CK-MB Creatine kinase isoenzymes, cTnI Cardiac troponin I, COPD Chronic obstructive disease, Dbil Direct bilirubin, ECMO Extracorporeal membrane oxygenation, FDP Fibrin degradation products, FIB Fibrinogen, ILD Interstitial lung disease, INR International normalized ratio, Glu Glucose, LTx Lung transplantation, Myo Myoglobin, NEUT Neutrophils, PT Prothrombin time, PTA Prothrombin time activity percentage, RBC Red blood cells, SCr serum creatinine, T2DM Type 2 diabetes mellitus, Tac Tacrolimus, Tbil Total bilirubin, TP Total protein, TT Thrombin time, UA Uric acid, WBC White blood cells Basic characteristics of study population in the training and validation sets Abbreviations AKI Acute kidney injury, ALB Albumin, ALT Alanine aminotransferase, APTT Activated partial thromboplastin time, ASA American Society of Anesthesiologists, AST Aspartate aminotransferase, BMI Body mass index, BO Bronchiolitis obliterans, CK-MB Creatine kinase isoenzymes, cTnI Cardiac troponin I, COPD Chronic obstructive disease, Dbil Direct bilirubin, ECMO Extracorporeal membrane oxygenation, FDP Fibrin degradation products, FIB Fibrinogen, ILD Interstitial lung disease, INR International normalized ratio, Glu Glucose, LTx Lung transplantation, Myo Myoglobin, NEUT Neutrophils, PT Prothrombin time, PTA Prothrombin time activity percentage, RBC Red blood cells, SCr serum creatinine, T2DM Type 2 diabetes mellitus, Tac Tacrolimus, Tbil Total bilirubin, TP Total protein, TT Thrombin time, UA Uric acid, WBC White blood cells The distribution of SNPs The Supplementary Table S1 provides a comprehensive list of all SNPs investigated in this study, arranged in sequential order from chromosome 1 to 22. A total of 153 SNPs were selected and subjected to thorough analysis. It is worth noting that all the analyzed SNPs demonstrated adherence to Hardy–Weinberg equilibrium, except for HHEX rs7923837, IL-10 rs1800871, STAT3 rs744166, ADAMTS9 rs4607103, CYP24A1 rs2296241, JAK2 rs12343867 and NR1l2 rs2276707. Consequently, these non-compliant SNPs were excluded from further investigations. Correlation between SNPs and persistent AKI We developed three genetic models (codominant, dominant, and recessive) to investigate the association between SNPs and persistent AKI. P values were calculated for three model of each SNP, and statistical significance was displayed in a Manhattan plot (Fig. 1). Eight SNPs exhibited significant associations with persistent AKI. The most significant models were dominant for ATP binding cassette subfamily B member 1 (ABCB1) rs1045642 (p = 0.020), codominant for Glucokinase (GCK)rs1799884 (p = 0.013), dominant for Transforming Growth Factor Beta 1 (TGFB1) rs1800469 (p = 0.037), recessive for HNF1 Homeobox B (HNF1B) rs757210 (p = 0.005), recessive for Interleukin 12B (IL-12B) rs6887695 (p = 0.006), recessive for interleukin-2 (IL-2) rs2069762 (p = 0.020), recessive for interleukin-7 receptor (IL-7R) rs1494558 (p = 0.034) and recessive for Interleukin 17A (IL-17A) rs2275913 (p = 0.035) (Supplementary Table S2). Due to the limited number of patients in specific genotype subgroups, subsequent analyses for GCK rs1799884 was performed using the dominant model (p = 0.026). Manhattan plot of candidate SNPs. Dots indicate the positions of candidate SNPs. The dotted red line is the bonferroni-corrected significance threshold Construction of prediction model for persistent AKI The influence of the clinical parameters on persistent AKI was assessed using a logistic regression model, and the estimated effects of parameters with a p-value less than 0.1 are presented in Supplementary Table S3. In univariate analysis, BMI, indication, Tbil and Tac initial concentration exhibited significant correlations with the incidence of persistent AKI. In multivariate analysis, we combined clinical factors with p < 0.1 along with eight SNPs which showed significance in genetic analysis to construct the final prediction model based on the minimum Akaike information criterion (AIC) value. This model included nine predictors: BMI, TT, Tac initial concentration, GCK rs1799884, IL-7R rs1494558, HNF1B rs757210, IL-12B rs6887695 and IL-2 rs2069762 and IL-17A rs2275913 (Table 2). Table 2Multivariate analysis for prediction of persistent AKI following LTxParametersBSEORCIZp-valueBMI< 18.5 kg/m2Ref18.5–23.9 kg/m21.2060.5393.3403.340 (1.216–10.28)2.2380.025> 23.9 kg/m21.3590.5703.8943.893 (1.324–12.64)2.3840.017TT > 21 s−2.1681.1800.1140.114 (0.005–0.817)−1.8380.066Tac initial concentration > 6.5 ng/mL0.6650.3861.9441.944 (0.917–4.186)1.7240.085rs757210 (GG + GA vs. AA)1.8790.8576.5486.547 (1.451–48.10)2.1910.028rs2069762 (AA + AC vs. CC)−1.8570.6650.1560.156 (0.038–0.545)−2.7900.005rs1494558 (AA + AG vs. GG)−1.1410.4490.3200.319 (0.127–0.752)−2.5390.011rs6887695 (CC + CG vs. GG)1.5120.5854.5364.535 (1.528–15.68)2.5860.010rs2275913 (GG + GA vs. AA)−0.7820.4740.4570.457 (0.175–1.141)−1.6490.099rs1799884 (AA + GA vs. GG)0.5310.3561.7011.701 (0.847–3.443)1.4900.136Abbreviations AKI Acute kidney injury, BMI Body mass index, CI Confidence interval, LTx Lung transplantation, Tac Tacrolimus, TT Thrombin time, WBC White blood cells Multivariate analysis for prediction of persistent AKI following LTx Abbreviations AKI Acute kidney injury, BMI Body mass index, CI Confidence interval, LTx Lung transplantation, Tac Tacrolimus, TT Thrombin time, WBC White blood cells Evaluation of the prediction model The receiver operating characteristic (ROC) curve analysis depicted in Fig. 2 illustrates the predictive model’s accuracy in both the training and validation set (A). In the training set, the nomogram demonstrated an area under the curve (AUC) of 0.781 (95%CI: 0.715–0.846), while in the validation set, it achieved an AUC of 0.698 (95%CI: 0.542–0.855), indicating robust predictive accuracy and discrimination capabilities. The model’s predictive performance and stability were further validated through 500 iterations of bootstrap internal sampling, with an AUC of 0.769 (95%CI: 0.700–0.838) and 0.746 (95%CI:0.599–0.893) in the training and validation set, respectively, as shown in Fig. 2B and C. Additionally, the model achieved an F1-score of 87%, with precision and recall rates of 63% and 87% respectively in the validation set. And it obtained an F1-score of 62%, with precision and recall rates of 54% and 72% respectively in the validation set. ROC curve analysis for A the training set and the validation set. B and C depicted the ROC curve analysis which was validated by 500 iterations of bootstrap internal sampling in the training set and validation set, respectively The Hosmer–Lemeshow test showed a high level of agreement between predicted and actual probabilities (training set, p = 0.967; validation set, p = 0.288). The calibration curves presented in Fig. 3 exhibited a good fit of the prediction model across both sets after 500 iterations of Bootstrap internal sampling. The decision curve analysis (DCA) in Fig. 4 demonstrates the clinical utility of the prediction model in both the training and validation sets. The nomogram model exhibited a favorable net benefit across a broad range, with threshold probabilities ranging from 5 to 75% for the training set and from 20 to 55% for the validation set. Calibration curves for model prediction of persistent AKI in A the training set and B the validation set. The dashed line represented the ideal model, the blue line represented the calibration of the model constructed, and the red line represented the calibration after conduction 500 iterations of bootstrap internal sampling Decision curve analysis for A the training set and B the validation set of the model sampled by Bootstrap 500 times To further assess the predictive performance of the model, we compared the model’s AUC with those of individual predictors, SNPs alone, and clinical parameters alone, as shown in Fig. 5. The AUCs for single predictors, SNPs alone and clinical parameters alone were consistently lower than the model’s AUC in both training and validation sets, highlighting the model’s enhanced predictive power. Comparison of the performance of the nomogram and single predictors in predicting persistent AKI using ROC curves in A the training set and B the validation set and comparison of the performance of the nomogram (Mod A) and SNPs alone (ModB), clinical predictors alone (ModC) in predicting persistent AKI using ROC curves in C the training set and D the validation set Nomogram for prediction of persistent AKI in LTx patients A nomogram, as depicted in Fig. 6, was presented to predict the risk of persistent AKI in LTx patients. The nomogram comprised nine predictors: BMI, TT, Tac initial concentration, GCK rs1799884 genotype, IL-7R rs1494558 genotype, HNF1B rs757210 genotype, IL-12B rs6887695 genotype, IL-2 rs2069762 genotype and IL-17A rs2275913 genotype. To illustrate the use of this nomogram, let’s consider a patient who underwent LTx at our center. His BMI was 24.0 kg/m2 and his preoperative blood laboratory tests revealed a TT value of 22 s. On POD3, his peripheral blood was collected for the first Tac concentration measurement and the result returned 8 ng/mL. Genetic testing showed an AG genotype for rs1799884, AA for rs1494558, GG for rs757210, GG for rs6887695, CC for rs2069762 and AA for rs2275913. Each predictor corresponded to scores of 63, 0, 31, 24, 0, 87, 0, 86 and 36 points respectively. Consequently, his total score amounted to 327 points indicating a calculated risk of persistent AKI at approximately 42% for this patient. Predictive nomogram for persistent AKI after LTx Discussion Main findings In this study, we conducted a comprehensive genetic analysis and identified several specific SNPs, including GCK rs1799884, IL-7R rs1494558, HNF1B rs757210, IL-12B rs6887695, IL-2 rs2069762 and IL-17A rs2275913, which may confer an increased risk for the development of persistent AKI following LTx. Importantly, our study has successfully developed a novel nomogram for predicting persistent AKI in the LTx population by integrating both genetic information and clinical parameters. The nomogram demonstrated excellent discrimination and calibration during internal validation, making it a useful tool in the personalized medicine approach to post-transplant care. Clinical significance The definition of persistent AKI was initially proposed by the Acute Disease Quality Initiative (ADQI) in 2017, defining it as AKI lasting for more than 48 h [6]. The incidence of persistent AKI varies depending on the patient cohort and the specific criteria used to define AKI. In our study, we observed a prevalence rate of 43.7% for persistent AKI. When applying the KDIGO criteria and utilizing a 7-day evaluation period to identify persistent AKI, Jing et al. reported a prevalence rate of 49.2%, while Fidalgo documented a frequency rate of 33.5%, which is consistent with our findings [5, 8]. The clinical significance of persistent AKI has been emphasized by several studies, suggesting that this condition more accurately reflects the true clinical manifestation of AKI, since it demonstrates a stronger association with mortality risk, regardless of the criteria used to define AKI [11, 12]. For instance, in patients undergoing cardiac surgery, AKI lasting seven days or more is associated with a markedly higher in-hospital mortality rate (15.3%) compared to shorter durations of 1–2 days (4.1%) [13, 14]. Furthermore, the analysis of 5-year survival rates reveals an elevated risk of mortality for individuals experiencing persistent AKI (hazard ratio (HR): 3.40) compared to those with transient episodes (HR: 1.66) [14, 15]. In the context of LTx, there is limited number of literatures addressing the issue of persistent AKI. For instance, Park et al. [16] reported significant differences in 1-year mortality rates between patients without AKI and those with persistent AKI. Jing et al. [5] found lower survival rates at both 30 days and 1 year among patients with persistent AKI compared to those with transient AKI or no AKI. Fidalgo et al. [8] demonstrated that persistent AKI had negative impact on 5-year survival rates. Our research also revealed that patients with persistent AKI exhibited significantly inferior 90-day survival rates when compared to both non-persistent AKI and no-AKI patients, thereby consolidating the robust impact of persistent AKI on post-LTx mortality (Supplementary Figure S1). Clinical risk factor analysis Consequently, timely identification of risk factors for persistent AKI is crucial to prevent or promptly implement clinical interventions, thus improving patients’ prognosis. In our study, we identified BMI, TT and Tac initial concentration as risk factors for persistent AKI. Notably, a lower BMI (< 18.5 kg/m2) was associated with a decreased risk of persistent AKI compared to individuals with normal or higher BMIs. For patients who were overweight or obesity, this correlation may be attributed to obesity’s effects on renal hemodynamics, which can increase renal plasma flow and filtration rates, predisposing individuals to a higher risk of renal injury [17]. Similarly, several studies have also reported an association between higher BMI and the development of AKI [18–20]. Continuous administration of Tac after transplantation could lead to persistent renal impairment due to its nephrotoxicity. This is believed to be caused by acute vascular and tubular damage resulting from increased vasoconstrictors (such as endothelin and thromboxane), activation of the renin-angiotensin system, and reduced vasodilators (including prostaglandin E2, prostacyclin, and nitric oxide) [21]. In the LTx population, Sikma et al. [22] revealed that AKI was associated with prior supratherapeutic whole-blood Tac trough concentrations in the first week (OR 1.66; 95% CI 1.20–2.31). Miano et al. [23] demonstrated a correlation between higher levels of Tac C0 and both the occurrence of AKI (HR: 1.54, 95%CI: 1.20–1.96) as well as its severity (OR: 1.29, 95%CI: 1.04–1.60). In our study, we also found that an initial Tac concentration > 6.5 ng/mL is associated with an increased risk for persistent AKI. Genetic implications Genetic factors have been implicated as potential contributors to the susceptibility and severity of AKI. In our study, we identified six SNPs that may increase the risk of persistent AKI. Some of these SNPs have previously been associated with renal impairment. The pro-inflammatory cytokine IL-17 plays a pivotal role in various processes, including but not limited to IL-17A [24]. Genetic variations in the genes encoding IL-17A have been investigated in the pathogenesis of rejection following solid organ transplantation, and previous studies have demonstrated a significant association between IL-17A rs2275913 polymorphism and impaired long-term kidney allograft function [25], as well as elevated long-term creatinine levels in kidney transplant recipients [26]. IL-7 is a pivotal cytokine in T-cell development, homeostasis, and the establishment of immunological memory [27]. Variants of the IL-7 receptor α-chain (IL-7RA) play a crucial role in regulating IL-7 availability. It has been reported that IL7R is significantly upregulated in diabetic kidney disease and exhibits a negative correlation with glomerular filtration rate [28]. Specifically, it has been demonstrated that IL-7RA rs1494558 tags haplotypes associated with susceptibility to autoimmune diseases [29]. Hahn et al. [30] reported an association between missense SNPs of IL-7R, specifically rs1494558, and susceptibility to nephrotic range proteinuria in childhood IgA nephropathy. GCK serves as a crucial regulator of insulin secretion and synthesis, playing an essential role in maintaining glucose homeostasis [31]. Wang et al. [32] reported that individuals carrying the A allele of GCK rs1799884 were at a higher risk for end-stage kidney disease among Chinese patients with type 2 diabetes. HNF1B, belonging to the homeobox-containing family of transcription factors, functions as a crucial transcription factor involved in early nephron development by activating segment-specific gene expression responsible for nephron development [33]. Mutations in HNF1B are associated with phenotypic abnormalities including kidney disease, diabetes [34], liver abnormalities, and other diseases [35]. Specifically, the SNP rs757210 has been associated with the common form of type 2 diabetes [36]. The IL-2 gene encodes cytokines that activates cytotoxic T cells and NK cells [37] which have been suggested to be involved in the development of diabetes mellitus [38]. Additionally, the SNP rs2069762 has been reported to be correlated with new-onset diabetes after transplantation [39]. Given that hyperglycemia is a well-established primary pathogenic factor in diabetes mellitus-induced nephropathy, it is plausible that these SNPs may contribute to kidney impairment through intracellular hyperglycemia-induced generation of reactive oxygen species [40]. Although IL-12B rs6887695 have not been previously associated with kidney injury, its known involvement in cytokine production regulation and autoimmune diseases implies a potential contribution to kidney impairment [41]. Clinical integration The ultimate objective of identifying risk factors and constructing the prediction nomogram is to enhance prognosis. Given the high prevalence of persistent AKI after LT, early recognition and therapeutic interventions to restrict its occurrence and progression are crucial in reducing morbidity and mortality rates. One suggestion is to optimize immunosuppressive therapy while limiting exposure to nephrotoxic agents. Close monitoring of CNI trough concentration is imperative for lung transplant recipients. Another important measure involves conducting genetic testing to identify patients at high risk. Integrating renal protection strategies with the utilization of these SNPs for developing early kidney risk identification and protection protocols represents a promising concept that necessitates further investigation. Innovations and limitations Despite the increasing number of prediction models for AKI in recent years, there has been limited research specifically focused on AKI in transplant patients [42–44]. Previous studies by Zeng et al. [45] and Kalisvaart et al. [46] have developed prediction models for post-transplant AKI using different algorithms and regression analyses, achieving AUC values of 0.817 and 0.70 respectively. However, none of these studies have specifically been done in LTx patients. Our research is the first to combine genetic data with clinical predictors to predict persistent AKI in LTx patients, resulting in a more comprehensive and predictive model for clinical outcomes. Our study has certain limitations. Firstly, in comparison to the existing literature on prediction model building in other fields, our sample size is relatively small. However, it should be noted that the availability of LTx populations is limited and valuable, posing challenges for data collection. Moreover, when compared to previously published articles focusing on similar LTx populations, our sample size can be considered relatively larger, thereby enhancing the reliability of our conclusions. Secondly, it is important to highlight that when applied to the validation set, our model yielded an AUC of 0.698 with a moderately wide 95% CI range. This observation may be attributed to the aforementioned limitation of a small sample size. Additionally, we did not validate the functions of the genes by conducting in vitro experiments which could help elucidating the precise role of these SNPs in AKI development. We acknowledge that our current study serves as an exploratory endeavor; therefore, larger multi-center studies are warranted to establish this model as a robust clinical standard. Conclusion In summary, our investigation has successfully identified six single nucleotide polymorphisms (SNPs): GCK rs1799884, IL-7R rs1494558, HNF1B rs757210, IL-12B rs6887695, IL-2 rs2069762 and IL-17A rs2275913, that may increase the susceptibility to persistent AKI following LTx. By incorporating these genetic variants with relevant clinical parameters, we have developed a user-friendly nomogram aimed at predicting the risk of persistent AKI within the LTx population. The model demonstrated excellent discrimination and calibration during internal validation. A significant advantage of our predictive model lies in its clinical utility as it provides clinicians with a valuable tool for identifying patients at high risk of persistent AKI and potentially reducing avoidable postoperative renal complications. This advancement not only enhances our understanding of the genetic predisposition to persistent AKI in the context of LTx but also contributes to a personalized approach to patient care. Supplementary Information Supplementary Material 1. Acknowledgments Clinical trial number Not applicable. Authors’ contributions DWW contributed to the concept and design of this study. DWW, WXX, and ZD were responsible for the statistical analysis and writing of the report. ZXB and CWQ assisted in the statistical analysis. LPM reviewed the article and provided critical feedback to improve and structure the report. All the authors have read and approved the final manuscript. Funding The study was financed with National Key Research and Development Program of China (2020YFC2005504), National High Level Hospital Clinical Research Funding (2023-NHLHCRF-YYPPLC-TJ-22), CAMS Innovation Fund for Medical Science (2021-I2M-1-049), National Natural Science Foundation Project (82273523), and Capital's Funds for Health Improvement and Research (CFH 2020-1-2031). Data availability The variation data reported in this paper have been deposited in the Genome Variation Map (GVM) in National Genomics Data Center, Beijing Institute of Genomics, Chinese Academy of Sciences and China National Center for Bioinformation. Data could be accessed on the following URL link: https://bigd.big.ac.cn/gvm/getProjectDetail?Project=GVM000841 [47, 48]. Declarations Ethics approval and consent to participate All procedures in this study were in accordance with the 1964 Helsinki declaration and its amendments, and was approved by the Ethics Committee of China-Japan Friendship Hospital in June, 2022 (No. 2022-KY-056-1). A waiver for informed consent was granted by the Ethics Committee of China-Japan Friendship Hospital. Consent for publication Not applicable. Competing interests The authors declare no competing interests. Footnotes Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Change history 2/25/2025 A Correction to this paper has been published: 10.1186/s12882-025-04036-z References Associated Data Supplementary Materials Supplementary Material 1. Data Availability Statement The variation data reported in this paper have been deposited in the Genome Variation Map (GVM) in National Genomics Data Center, Beijing Institute of Genomics, Chinese Academy of Sciences and China National Center for Bioinformation. Data could be accessed on the following URL link: https://bigd.big.ac.cn/gvm/getProjectDetail?Project=GVM000841 [47, 48].