ABSTRACT
@#Objective To explore the application value of machine learning models in predicting postoperative survival of patients with thoracic squamous esophageal cancer. Methods The clinical data of 369 patients with thoracic esophageal squamous carcinoma who underwent radical esophageal cancer surgery at the Department of Thoracic Surgery of Northern Jiangsu People's Hospital from January 2014 to September 2015 were retrospectively analyzed. There were 279 (75.6%) males and 90 (24.4%) females aged 41-78 years. The patients were randomly divided into a training set (259 patients) and a test set (110 patients) with a ratio of 7 : 3. Variable screening was performed by selecting the best subset of features. Six machine learning models were constructed on this basis and validated in an independent test set. The performance of the models' predictions was evaluated by area under the curve (AUC), accuracy and logarithmic loss, and the fit of the models was reflected by calibration curves. The best model was selected as the final model. Risk stratification was performed using X-tile, and survival analysis was performed using the Kaplan-Meier method with log-rank test. Results The 5-year postoperative survival rate of the patients was 67.5%. All clinicopathological characteristics of patients between the two groups in the training and test sets were not statistically different (P>0.05). A total of seven variables, including hypertension, history of smoking, history of alcohol consumption, degree of tissue differentiation, pN stage, vascular invasion and nerve invasion, were included for modelling. The AUC values for each model in the independent test set were: decision tree (AUC=0.796), support vector machine (AUC=0.829), random forest (AUC=0.831), logistic regression (AUC=0.838), gradient boosting machine (AUC=0.846), and XGBoost (AUC=0.853). The XGBoost model was finally selected as the best model, and risk stratification was performed on the training and test sets. Patients in the training and test sets were divided into a low risk group, an intermediate risk group and a high risk group, respectively. In both data sets, the differences in surgical prognosis among three groups were statistically significant (P<0.001). Conclusion Machine learning models have high value in predicting postoperative prognosis of thoracic squamous esophageal cancer. The XGBoost model outperforms common machine learning methods in predicting 5-year survival of patients with thoracic squamous esophageal cancer, and it has high utility and reliability.
ABSTRACT
Objective@#To investigate the current status and real performance of the detection of RUNX1-RUNX1T1 fusion transcript levels and WT1 transcript levels in China through interlaboratory comparison.@*Methods@#Peking University People’s Hospital (PKUPH) prepared the samples for comparison. That is, the fresh RUNX1-RUNX1T1 positive (+) bone morrow nucleated cells were serially diluted with RUNX1-RUNX1T1 negative (-) nucleated cells from different patients. Totally 23 sets with 14 different samples per set were prepared. TRIzol reagent was added in each tube and thoroughly mixed with cells for homogenization. Each laboratory simultaneously tested RUNX1-RUNX1T1 and WT1 transcript levels of one set of samples by real-time quantitative PCR method. All transcript levels were reported as the percentage of RUNX1-RUNX1T1 or WT1 transcript copies/ABL copies. Spearman correlation coefficient between the reported transcript levels of each participated laboratory and those of PKUPH was calculated.@*Results@#①RUNX1-RUNX1T1 comparison: 9 samples were (+) and 5 were (-) , the false negative and positive rates of the 20 participated laboratories were 0 (0/180) and 5% (5/100) , respectively. The reported transcript levels of all 9 positive samples were different among laboratories. The median reported transcript levels of 9 positive samples were from 0.060% to 176.7%, which covered 3.5-log. The ratios of each sample’s highest to the lowest reported transcript levels were from 5.5 to 12.3 (one result which obviously deviated from other laboratories’ results was not included) , 85% (17/20) of the laboratories had correlation coefficient ≥0.98. ②WT1 comparison: The median reported transcript levels of all 14 samples were from 0.17% to 67.6%, which covered 2.6-log. The ratios of each sample’s highest to the lowest reported transcript levels were from 5.3-13.7, 62% (13/21) of the laboratories had correlation coefficient ≥0.98. ③ The relative relationship of the reported RUNX1-RUNX1T1 transcript levels between the participants and PKUPH was not always consistent with that of WT1 transcript levels. Both RUNX1-RUNX1T1 and WT1 transcript levels from 2 and 7 laboratories were individually lower than and higher than those of PKUPH, whereas for the rest 11 laboratories, one transcript level was higher than and the other was lower than that of PKUPH.@*Conclusion@#The reported RUNX1-RUNX1T1 and WT1 transcript levels were different among laboratories for the same sample. Most of the participated laboratories reported highly consistent result with that of PKUPH. The relationship between laboratories of the different transcript levels may not be the same.
ABSTRACT
Objective To explore the relationship between WT1 and prognosis of patients with acute myeloid leukemia (AML) or acute lymphoblastic leukemia (ALL), and to evaluate the possibility of WT1 as a potential marker for monitoring the minimal residual disease (MRD). Methods Bone marrow mononuclear cells from 58 patients with primary AML, 32 patients with primary ALL, 40 patients with AML-complete remission (CR), 28 patients with ALL-CR and 31 patients with trilineage hyperplasia (control group) were collected. Real-time fluorescent quantitative PCR method was used to detect the expression of WT1 in all patients. The expression threshold of WT1 in each group was established. WT1 copy number/ABL copy number ratio×100%denotes the relative expression level of WT1 gene. Results Median relative expression level of WT1 in the control patients was much lower than that in primary AML patients [0.026%(0-0.240%) vs. 20.880 % (3.550 %-48.500 %), Z=-7.74, P20.880 %) was 60.7 % (17/28), while the CR rate was 76.7 % (23/30) in those with lower WT1 expression. WT1 expression was increased dramatically in recurrent AML patients. Relative expression level of WT1 was significantly higher in primary ALL patients [0.350 % (0.021 %-10.780 %)] compared with that in the control group Z=-2.58, P<0.05. There was no significant difference in relative expression level of WT1 between ALL and ALL-CR patients [0.038 % (0-2.800 %), P=0.065]. Conclusion WT1 expression level in AML patients is relatively high, which could be used as an effective index of prognosis evaluation and MRD monitoring for AML patients, but not for ALL patients.
ABSTRACT
<p><b>OBJECTIVE</b>To revalidate the conversion factor（CF）for the conversion of BCR-ABL （P210）transcript levels to the international scale（BCR- ABLIS）in chronic myeloid leukemia（CML） which validated before.</p><p><b>METHODS</b>Peking University People's Hospital（PKUPH）prepared the exchange samples for revalidation of CFs of 15 laboratories which validated nine or eighteen months ago. The fresh BCR-ABL（P210）（+）bone morrow or peripheral blood nucleated cells were diluted with BCR-ABL （P210）（-）cells to achieve different BCR- ABL levels, totally 16 sets and 24 samples per set were prepared. TRIzol reagent was added in each tube. Each laboratory tested BCR-ABL transcript levels of one set of samples. Agreement between BCR-ABLIS of each laboratory and PKUPH was assessed by the Bland- Altman method. For laboratories which did not meet the criteria of revalidation, linear regression equation was derived after the samples with maximum BCR-ABL deviation were removed until R²>0.98, then new CF was calculated.</p><p><b>RESULTS</b>10 laboratories met the revalidation criteria with both bias within ±1.4 fold and 95% limits of agreement within ±6 folds, and their CFs still could be used for accurately conversion of BCR-ABLIS. New CFs were recalculated as of 1.8-6.3 folds of their previous CFs in 5 laboratories not met the criteria.</p><p><b>CONCLUSION</b>Revalidation of CF by sample exchange among laboratories was necessary for accurate and continuous application of BCR-ABLIS, which not only tested the validity of CF acquired before but also calculated new available CFs for those with invalid CFs.</p>
Subject(s)
Humans , Bone Marrow Cells , Fusion Proteins, bcr-abl , Genetics , Leukemia, Myelogenous, Chronic, BCR-ABL Positive , Diagnosis , GeneticsABSTRACT
Objective Estahlished the method to detect different transcripts of EVI1 gene expression with quantitative PCR and study the expression patterns of EVI1 gene in different leukemia groups to investigate the association between EVI1 gene expression and the incidence and prognosis of leukemia.Methods 60 cases acute myeloid leukemia (AML) and 9 cases normal control were detected in the study,37 cases were male and 32 cases were female,age 10-70 years,median age 42 years,M3 36 cases,M2 16 cases and M4 8 cases according to FAB classification criteria,control samples of nine cases were normal healthy people.Using the quantitative PCR (Taq Man probe) to detect the expression of different transcripts of EVI1 gene.The t test was used to detect the expression difference among different leukemia groups.Results ABL gene was used as internal reference,relative changes of EVI1 gene expression level were detected by EVI1/ABL.In all the control patients,EVI1 gene of different transcription of this expression were detected,expression level of EVI1 gene different transcription was significant with the difference (P ＜ 0.05),transcription 2 and 5 (the same primers) were the lowest,followed for transcription 1 and 6,expression of transcription 3 was the highest.The expression levels of transcripts 2 and 5,1,6,3 were nagative,0.005,0.050 and 0.512 respectively in healthy control samples.In addition,the EVI1 gene expression was negatively correlated with expression of the fusion gene AML-ETO and CBFB-MYH11 in AML.Conclusion The study established a stable,fast and accurate method to detect the expression of EVI1 gene.