Influencing factors on false positive rates when classifying tumor cell line response to drug treatment
Informed selection of drug candidates for laboratory experimentation provides an efficient means of identifying suitable anti-cancer treatments. The advancement of artificial intelligence has led to the development of computational models to predict cancer cell line response to drug treatment. It is important to analyze the false positive rate (FPR) of the models, to increase the number of effective treatments identified and to minimize unnecessary laboratory experimentation. Such analysis will also aid in identifying drugs or cancer types that require more data collection to improve model predictions. This work uses an attention based neural network classification model to identify responsive/non-responsive drug treatments across multiple types of cancer cell lines. Two data filtering techniques have been applied to generate 10 data subsets, including removing samples for which dose response curves are poorly fitted and removing samples whose area under the dose response curve (AUC) values are marginal around 0.5 from the training set. One hundred trials of 10-fold cross-validation analysis is performed to test the model prediction performance on all the data subsets and the subset with the best model prediction performance is selected for further analysis. Several error analysis metrics such as the false positive rate (FPR), and the prediction uncertainty are evaluated, and the results are summarized by cancer type and drug mechanism of action (MoA) category. The FPR of cancer type spans between 0.262 and 0.5189, while that of drug MoA category spans almost the full range of [0, 1]. This study identifies cancer types and drug MoAs with high FPRs. Additional drug screening data of these cancer and drug categories may improve response modeling. Our results also demonstrate that the two data filtering approaches help improve the drug response prediction performance.
Code (0)
등록된 구현이 없습니다.
Tasks
Drug Response PredictionSimilar Papers 제목 키워드 기반
Robust Knockoffs for Controlling False Discoveries With an Application to Bond Recovery Rates
We address challenges in variable selection with highly correlated data that are frequently present in finance, economics, but also in complex natural systems as e.g. weather. We develop a robustified version of the knoc…
Variable SelectionAn Insight into Security Code Review with LLMs: Capabilities, Obstacles, and Influential Factors
Security code review is a time-consuming and labor-intensive process typically requiring integration with automated security defect detection tools. However, existing security analysis tools struggle with poor generaliza…
Defect DetectionMultivariate outlier detection based on a robust Mahalanobis distance with shrinkage estimators
A collection of robust Mahalanobis distances for multivariate outlier detection is proposed, based on the notion of shrinkage. Robust intensity and scaling factors are optimally estimated to define the shrinkage. Some p…
Outlier DetectionSystematic Categorization of Influencing Factors on Radar-Based Perception to Facilitate Complex Real-World Data Evaluation
For the assessment of machine perception for automated driving it is important to understand the influence of certain environment factors on the sensors used. Especially when investigating large amounts of real-world dat…
Sensor ModelingInfluence of uncertainty estimation techniques on false-positive reduction in liver lesion detection
Deep learning techniques show success in detecting objects in medical images, but still suffer from false-positive predictions that may hinder accurate diagnosis. The estimated uncertainty of the neural network output ha…
Lesion Detection