Background: Cancer cell lines are frequently used in research as in-vitro tumor models. Genomic data and large-scale drug screening have accelerated the right drug selection for cancer patients. Accuracy in drug response prediction is crucial for success. Due to data-type diversity and big data volume, few methods can integrative and efficiently find the principal low-dimensional manifold of the high-dimensional cancer multi-omics data to predict drug response in precision medicine. Method: A novelty k-means Ensemble Support Vector Regression (kESVR) is developed to predict each drug response values for single patient based on cell-line gene expression data. The kESVR is a blend of supervised and unsupervised learning methods and is entirely data driven. It utilizes embedded clustering (Principal Component Analysis and k-means clustering) and local regression (Support Vector Regression) to predict drug response and obtain the global pattern while overcoming missing data and outliers’ noise. Results: We compared the efficiency and accuracy of kESVR to 4 standard machine learning regression models: (1) simple linear regression, (2) support vector regression (3) random forest (quantile regression forest) and (4) back propagation neural network. Our results, which based on drug response across 610 cancer cells from Cancer Cell Line Encyclopedia (CCLE) and Cancer Therapeutics Response Portal (CTRP v2), proved to have the highest accuracy (smallest mean squared error (MSE) measure). We next compared kESVR with existing 17 drug response prediction models based a varied range of methods such as regression, Bayesian inference, matrix factorization and deep learning. After ranking the 18 models based on their accuracy of prediction, kESVR ranks first (best performing) in majority (74%) of the time. As for the remaining (26%) cases, kESVR still ranked in the top five performing models. Conclusion: In this paper we introduce a novel model (kESVR) for drug response prediction using high dimensional cell-line gene expression data. This model outperforms current existing prediction models in terms of prediction accuracy and speed and overcomes overfitting. This can be used in future to develop a robust drug response prediction system for cancer patients using the cancer cell-lines guidance and multi-omics data.

Background: Large-scale screening of drug sensitivity on cancer cell models can mimic in vivo cellular behavior providing wider scope for biological research on cancer. Since the therapeutic effect of a single drug or drug combination depends on the individual patient’s genome characteristics and cancer cells integration reaction, the identification of an effective agent in an in vitro model by using large number of cancer cell models is a promising approach for the development of targeted treatments. Precision cancer medicine is to select the most appropriate treatment or treatments for an individual patient. However, it still lacks the tools to bridge the gap between conventional in vitro cancer cell models and clinical patient response to inhibitors. Methods: An optimal two-layer decision system model is developed to identify the cancer cells that most closely resemble an individual tumor for optimum therapeutic interventions in precision cancer medicine. Accordingly, an optimal grid parameters selection is designed to seek the highest accordance for treatment selection to the patient’s preference for drug response and in vitro cancer cell drug screening. The optimal two-layer decision system model overcomes the challenge of heterology data comparison between the tumor and the cancer cells, as well as between the continual variation of drug responses in vitro and the discrete ones in clinical practice. We simulated the model accuracy using 681 cancer cells’ mRNA and associated 481 drug screenings and validated our results on 315 breast cancer patients drug selection across seven drugs (docetaxel, doxorubicin, fluorouracil, paclitaxel, tamoxifen, cyclophosphamide, lapitinib). Results: Comparing with the real response of a drug in clinical patients, the novel model obtained an overall average accordance over 90.8% across the seven drugs. At the same time, the optimal cancer cells and the associated optimal therapeutic efficacy of cancer drugs are recommended. The novel optimal two-layer decision system model was used on 1097 patients with breast cancer in guiding precision medicine for a recommendation of their optimal cancer cells (30 cancer cells) and associated efficacy of certain cancer drugs. Our model can detect the most similar cancer cells for each individual patient. Conclusion: A successful clinical translation model (optimal two-layer decision system model) was developed to bridge in-vitro basic science to clinical practice in a therapeutic intervention application for the first time. The novel tool kills two birds with one stone. It can help basic science to seek optimal cancer cell models for an individual tumor, while prioritizing clinical drugs’ recommendations in practice. Tool associated platform website: We extended the breast cancer research to 32 more types of cancers across 45 therapy predictions.

Preeclampsia is a complex and multifactorial disease and a leading cause of death in pregnancy with no current effective treatment strategies, especially severe preeclampsia. A new general regression model with biclustering scalability is developed to quantify causal relationship from genes to gene products to clinical phenotypes among multiple datasets. It can identify those key genes and relationships between gene expression and clinical phenotype, which will be interpreted further, for molecular mechanism disclosure and effective drug development by pathway enrichment analysis. We observed pattern variation of gene expression relating to clinical phenotype among three preeclampsia datasets in parallel. Eighty-one genes are recommended as potential druggable targets for severe preeclampsia by their highly correlation between gene expression variation with preeclampsia clinical phenotypes in hypertension, HELLP (Hemolysis, Elevated Liver enzymes and Low Platelets) syndrome, proteinuria, and gestation week. Genes ENG, Flt and NEU1 and NEU2 show overexpression in severe preeclampsia in attenuating hypertension and proteinuria in multiple datasets of preeclampsia. Those genes are involved in the angiogenic factors and vascular molecular function. Treatment with either Flt or NEU1 have exciting clinical implications, and likely will transform the detection and treatment for patients with severe preeclampsia. The novel network mathematical models can quantify a causal relationship from genes to gene products to clinical phenotypes, and they can model causal influences of genes that are either monomorphic or polymorphic among heterology datasets, which is suggesting that key gene expression variation may be more likely to influence complex clinical phenotype variation in preeclampsia among multiple datasets.