Abstract 87P
Background
ORIENT-3 was an open-label, multicenter, randomized controlled phase III study that recruited patients with stage ⅢB/ⅢC/Ⅳ squamous-cell non-small cell lung cancer (sqNSCLC) after failure with first-line platinum-based chemotherapy, which had response records and RNA data. This study used the RNA data of ORIENT-3 study to construct an artificial neural network (ANN) system to predict the response to sintilimab for patients with sqNSCLC.
Methods
RNA data were normalized based on the expression of the sum of two house keeping genes (ACTB and GAPDH), then were converted logarithmically. Patients treated with sintilimab were randomly divided into training cohort (70%) and test cohort (30%). Receiver operating characteristic curve (ROC) were used in training cohort to select genes, and the top 30 significant (p values < 0.05, area under the curve [AUC] of ROC > 0.75 or < 0.25, ranked by random forest method) genes were used to construct ANN models. Three hundred times of three-fold cross validation were performed to obtain 900 ANN models and test the accuracy of them. The final predicted values of the ANN system (900 ANN models) were equal to the weighted average of output values of all ANN models. Finally, the accuracy of the ANN system was tested in the test cohort. In addition, according linear models’ system was also built and was compared with the ANN system.
Results
In this study, 59 sintilimab-treated patients with RNA data were used to construct the ANN system. Among these patients, 14 had response (complete or partial response). After selection, 30 genes, includings WDR47, CCL13, WIF1, etc., were used to build three-layer ANN models. Each ANN model had 30 input nodes, four hidden nodes and one output node. The AUC was 1.00 (95% confidence interval [CI] 1.00-1.00) in the training cohort while 0.93 (95%CI 0.78-1.00) in the test cohort. In terms of the linear system, the AUC was 0.90 (95%CI 0.81-1.00) in the training cohort while 0.45 (95%CI 0.00-0.92) in the test cohort.
Conclusions
The ANN system showed a potential value in predicting the response to sintilimab for patients with sqNSCLC.
Legal entity responsible for the study
The authors.
Funding
Has not received any funding.
Disclosure
All authors have declared no conflicts of interest.