ITNR: Inversion Transformer-based Neural Ranking for cancer drug recommendations

Personalized drug response prediction is an approach for tailoring effective therapeutic strategies for patients based on their tumors' genomic characterization. While machine learning methods are widely employed in the literature, they often struggle to capture drug-cell line relations across various cell lines. In addressing this challenge, our study introduces a novel listwise Learning-to-Rank (LTR) model named Inversion Transformer-based Neural Ranking (ITNR). ITNR utilizes genomic features and a transformer architecture to decipher functional relationships and construct models that can predict patient-specific drug responses. Our experiments were conducted on three major drug response data sets, showing that ITNR reliably and consistently outperforms state-of-the-art LTR models.

Optimized models and deep learning methods for drug response prediction in cancer treatments: a review

Recent advancements in deep learning (DL) have played a crucial role in aiding experts to develop personalized healthcare services, particularly in drug response prediction (DRP) for cancer patients. The DL's techniques contribution to this field is significant, and they have proven indispensable in the medical field. This review aims to analyze the diverse effectiveness of various DL models in making these predictions, drawing on research published from 2017 to 2023. We utilized the VOS-Viewer 1.6.18 software to create a word cloud from the titles and abstracts of the selected studies. This study offers insights into the focus areas within DL models used for drug response. The word cloud revealed a strong link between certain keywords and grouped themes, highlighting terms such as deep learning, machine learning, precision medicine, precision oncology, drug response prediction, and personalized medicine. In order to achieve an advance in DRP using DL, the researchers need to work on enhancing the models' generalizability and interoperability. It is also crucial to develop models that not only accurately represent various architectures but also simplify these architectures, balancing the complexity with the predictive capabilities. In the future, researchers should try to combine methods that make DL models easier to understand; this will make DRP reviews more open and help doctors trust the decisions made by DL models in cancer DRP.

Uncovering the subtype-specific disease module and the development of drug response prediction models for glioma

The poor prognosis of glioma patients brought attention to the need for effective therapeutic approaches for precision therapy. Here, we deployed algorithms relying on network medicine and artificial intelligence to design the framework for subtype-specific target identification and drug response prediction in glioma. We identified the driver mutations that were differentially expressed in each subtype of lower-grade glioma and glioblastoma multiforme and were linked to cancer-specific processes. Driver mutations that were differentially expressed were also subjected to subtype-specific disease module identification. The drugs from the drug bank database were retrieved to target these disease modules. However, the efficacy of anticancer drugs depends on the molecular profile of the cancer and varies among cancer patients due to intratumor heterogeneity. Hence, we developed a deep-learning-based drug response prediction framework using the experimental drug screening data. Models for 30 drugs that can target the disease module were developed, where drug response measured by IC50 was considered a response and gene expression and mutation data were considered predictor variables. The model construction consists of three steps: feature selection, data integration, and classification. We observed the consistent performance of the models in training, test, and validation datasets. Drug responses were predicted for particular cell lines derived from distinct subtypes of gliomas. We found that subtypes of gliomas respond differently to the drug, highlighting the importance of subtype-specific drug response prediction. Therefore, the development of personalized therapy by integrating network medicine and a deep learning-based approach can lead to cancer-specific treatment and improved patient care.

DeepAEG: a model for predicting cancer drug response based on data enhancement and edge-collaborative update strategies

MOTIVATION

The prediction of cancer drug response is a challenging subject in modern personalized cancer therapy due to the uncertainty of drug efficacy and the heterogeneity of patients. It has been shown that the characteristics of the drug itself and the genomic characteristics of the patient can greatly influence the results of cancer drug response. Therefore, accurate, efficient, and comprehensive methods for drug feature extraction and genomics integration are crucial to improve the prediction accuracy.

RESULTS

Accurate prediction of cancer drug response is vital for guiding the design of anticancer drugs. In this study, we propose an end-to-end deep learning model named DeepAEG which is based on a complete-graph update mode to predict IC50. Specifically, we integrate an edge update mechanism on the basis of a hybrid graph convolutional network to comprehensively learn the potential high-dimensional representation of topological structures in drugs, including atomic characteristics and chemical bond information. Additionally, we present a novel approach for enhancing simplified molecular input line entry specification data by employing sequence recombination to eliminate the defect of single sequence representation of drug molecules. Our extensive experiments show that DeepAEG outperforms other existing methods across multiple evaluation parameters in multiple test sets. Furthermore, we identify several potential anticancer agents, including bortezomib, which has proven to be an effective clinical treatment option. Our results highlight the potential value of DeepAEG in guiding the design of specific cancer treatment regimens.

Advances towards the use of gastrointestinal tumor patient-derived organoids as a therapeutic decision-making tool

In December 2022 the US Food and Drug Administration (FDA) removed the requirement that drugs in development must undergo animal testing before clinical evaluation, a declaration that now demands the establishment and verification of ex vivo preclinical models that closely represent tumor complexity and that can predict therapeutic response. Fortunately, the emergence of patient-derived organoid (PDOs) culture has enabled the ex vivo mimicking of the pathophysiology of human tumors with the reassembly of tissue-specific features. These features include histopathological variability, molecular expression profiles, genetic and cellular heterogeneity of parental tissue, and furthermore growing evidence suggests the ability to predict patient therapeutic response. Concentrating on the highly lethal and heterogeneous gastrointestinal (GI) tumors, herein we present the state-of-the-art and the current methodology of PDOs. We highlight the potential additions, improvements and testing required to allow the ex vivo of study the tumor microenvironment, as well as offering commentary on the predictive value of clinical response to treatments such as chemotherapy and immunotherapy.

HMM-GDAN: Hybrid multi-view and multi-scale graph duplex-attention networks for drug response prediction in cancer

Precision medicine is devoted to discovering personalized therapy for complex and difficult diseases like cancer. Many machine learning approaches have been developed for drug response prediction towards precision medicine. Notwithstanding, genetic profiles based multi-view graph learning schemes have not yet been explored for drug response prediction in previous works. Furthermore, multi-scale latent feature fusion is not considered sufficiently in the existing frameworks of graph neural networks (GNNs). Previous works on drug response prediction mainly depend on sequence data or single-view graph data. In this paper, we propose to construct multi-view graph by means of multi-omics data and STRING protein-protein association data, and develop a new architecture of GNNs for drug response prediction in cancer. Specifically, we propose hybrid multi-view and multi-scale graph duplex-attention networks (HMM-GDAN), in which both multi-view self-attention mechanism and view-level attention mechanism are devised to capture the complementary information of views and emphasize on the importance of each view collaboratively, and rich multi-scale features are constructed and integrated to further form high-level representations for better prediction. Experiments on GDSC2 dataset verify the superiority of the proposed HMM-GDAN when compared with state-of-the-art baselines. The effectiveness of multi-view and multi-scale strategies is demonstrated by the ablation study.

Discovering the mechanism of action of drugs with a sparse explainable network

BACKGROUND

Although Deep Neural Networks (DDNs) have been successful in predicting the efficacy of cancer drugs, the lack of explainability in their decision-making process is a significant challenge. Previous research proposed mimicking the Gene Ontology structure to allow for interpretation of each neuron in the network. However, these previous approaches require huge amount of GPU resources and hinder its extension to genome-wide models.

METHODS

We developed SparseGO, a sparse and interpretable neural network, for predicting drug response in cancer cell lines and their Mechanism of Action (MoA). To ensure model generalization, we trained it on multiple datasets and evaluated its performance using three cross-validation schemes. Its efficiency allows it to be used with gene expression. In addition, SparseGO integrates an eXplainable Artificial Intelligence (XAI) technique, DeepLIFT, with Support Vector Machines to computationally discover the MoA of drugs.

FINDINGS

SparseGO's sparse implementation significantly reduced GPU memory usage and training speed compared to other methods, allowing it to process gene expression instead of mutations as input data. SparseGO using expression improved the accuracy and enabled its use on drug repositioning. Furthermore, gene expression allows the prediction of MoA using 265 drugs to train it. It was validated on understudied drugs such as parbendazole and PD153035.

INTERPRETATION

SparseGO is an effective XAI method for predicting, but more importantly, understanding drug response.

FUNDING

The Accelerator Award Programme funded by Cancer Research UK [C355/A26819], Fundación Científica de la AECC and Fondazione AIRC, Project PIBA_2020_1_0055 funded by the Basque Government and the Synlethal Project (RETOS Investigacion, Spanish Government).

Predicting heterogeneity in clone-specific therapeutic vulnerabilities using single-cell transcriptomic signatures

While understanding molecular heterogeneity across patients underpins precision oncology, there is increasing appreciation for taking intra-tumor heterogeneity into account. Based on large-scale analysis of cancer omics datasets, we highlight the importance of intra-tumor transcriptomic heterogeneity (ITTH) for predicting clinical outcomes. Leveraging single-cell RNA-seq (scRNA-seq) with a recommender system (CaDRReS-Sc), we show that heterogeneous gene-expression signatures can predict drug response with high accuracy (80%). Using patient-proximal cell lines, we established the validity of CaDRReS-Sc's monotherapy (Pearson r>0.6) and combinatorial predictions targeting clone-specific vulnerabilities (>10% improvement). Applying CaDRReS-Sc to rapidly expanding scRNA-seq compendiums can serve as in silico screen to accelerate drug-repurposing studies. Availability: https://github.com/CSB5/CaDRReS-Sc .

Mucosal IL23A expression predicts the response to Ustekinumab in inflammatory bowel disease

BACKGROUND

Biologics against tumor necrosis factor-α (TNF) and the p40 subunit of interleukin (IL)-12 and IL-23 are increasingly used in inflammatory bowel disease (IBD) treatment. However, information on response prediction to these agents is limited. Thus, we aimed to identify factors for IBD treatment response prediction.

METHODS

We conducted a retrospective study in 33 IBD subjects for anti-TNF and a prospective study of 23 IBD and 11 non-IBD subjects for ustekinumab (UST). Mucosal biopsy specimens were obtained before treatment with biologics. The expression of 18 immune-related genes encoding representative cytokines and transcription factors was analyzed by quantitative polymerase chain reaction.

RESULTS

There was no difference between the treatment-resistant and -sensitive groups with regard to clinical characteristics. A higher expression of oncostatin M (OSM) and its receptor OSMR in the intestinal mucosa was most strongly associated with anti-TNF resistance, whereas lower IL23A expression was most strongly associated with UST resistance. In addition to the absolute expression levels of genes, concordant or discordant expression patterns of particular gene sets were associated with treatment sensitivity and resistance.

CONCLUSIONS

The association of anti-TNF resistance and mucosal OSM and OSMR expression was consistent with the results of a previous study in a European cohort. Our observation that IBD subjects with higher mucosal IL23A expression were more likely to achieve remission by UST has not been previously reported. The response to biologics may thus be predicted in IBD patients through the analysis of mucosal gene expression levels and patterns.

A Molecular Signature Response Classifier to Predict Inadequate Response to Tumor Necrosis Factor-α Inhibitors: The NETWORK-004 Prospective Observational Study

Introduction

Timely matching of patients to beneficial targeted therapy is an unmet need in rheumatoid arthritis (RA). A molecular signature response classifier (MSRC) that predicts which patients with RA are unlikely to respond to tumor necrosis factor-α inhibitor (TNFi) therapy would have wide clinical utility.

Methods

The protein–protein interaction map specific to the rheumatoid arthritis pathophysiology and gene expression data in blood patient samples was used to discover a molecular signature of non-response to TNFi therapy. Inadequate response predictions were validated in blood samples from the CERTAIN cohort and a multicenter blinded prospective observational clinical study (NETWORK-004) among 391 targeted therapy-naïve and 113 TNFi-exposed patient samples. The primary endpoint evaluated the ability of the MSRC to identify patients who inadequately responded to TNFi therapy at 6 months according to ACR50. Additional endpoints evaluated the prediction of inadequate response at 3 and 6 months by ACR70, DAS28-CRP, and CDAI.

Results

The 23-feature molecular signature considers pathways upstream and downstream of TNFα involvement in RA pathophysiology. Predictive performance was consistent between the CERTAIN cohort and NETWORK-004 study. The NETWORK-004 study met primary and secondary endpoints. A molecular signature of non-response was detected in 45% of targeted therapy-naïve patients. The MSRC had an area under the curve (AUC) of 0.64 and patients were unlikely to adequately respond to TNFi therapy according to ACR50 at 6 months with an odds ratio of 4.1 (95% confidence interval 2.0–8.3, p value 0.0001). Odds ratios (3.4–8.8) were significant (p value < 0.01) for additional endpoints at 3 and 6 months, with AUC values up to 0.74. Among TNFi-exposed patients, the MSRC had an AUC of up to 0.83 and was associated with significant odds ratios of 3.3–26.6 by ACR, DAS28-CRP, and CDAI metrics.

Conclusion

The MSRC stratifies patients according to likelihood of inadequate response to TNFi therapy and provides patient-specific data to guide therapy choice in RA for targeted therapy-naïve and TNFi-exposed patients.

Supplementary Information

The online version contains supplementary material available at 10.1007/s40744-021-00330-y.

A blood-based molecular signature response classifier (MSRC) integrating next-generation RNA sequencing data with clinical features predicts the likelihood that a patient with rheumatoid arthritis will have an inadequate response to TNFi therapy. Treatment selection guided by test results, with likely inadequate responders appropriately redirected to a different therapy, could improve response rates to TNFi therapies, generate healthcare cost savings, and increase rheumatologists’ confidence in prescribing decisions and altered treatment choices. The MSRC described in this study predicts the likelihood of inadequate response to TNFi therapies among targeted therapy-naïve and TNFi-exposed patients in a multicenter, 24-week blinded prospective clinical study: NETWORK-004. Patients with a molecular signature of non-response are less likely to have an adequate response to TNFi therapies than those patients lacking the signature according to ACR50, ACR70, CDAI, and DAS28-CRP with significant odds ratios of 3.4–8.8 for targeted therapy-naïve patients and 3.3–26.6 for TNFi-exposed patients. This MSRC provides a solution to the long-standing need for precision medicine tools to predict drug response in rheumatoid arthritis—a heterogeneous and progressive disease with an abundance of therapeutic options. These data validate the performance of the MSRC in a blinded prospective clinical study of targeted therapy-naïve and TNFi therapy-exposed patients.

Super.FELT: supervised feature extraction learning using triplet loss for drug response prediction with multi-omics data

BACKGROUND

Predicting the drug response of a patient is important for precision oncology. In recent studies, multi-omics data have been used to improve the prediction accuracy of drug response. Although multi-omics data are good resources for drug response prediction, the large dimension of data tends to hinder performance improvement. In this study, we aimed to develop a new method, which can effectively reduce the large dimension of data, based on the supervised deep learning model for predicting drug response.

RESULTS

We proposed a novel method called Supervised Feature Extraction Learning using Triplet loss (Super.FELT) for drug response prediction. Super.FELT consists of three stages, namely, feature selection, feature encoding using a supervised method, and binary classification of drug response (sensitive or resistant). We used multi-omics data including mutation, copy number aberration, and gene expression, and these were obtained from cell lines [Genomics of Drug Sensitivity in Cancer (GDSC), Cancer Cell Line Encyclopedia (CCLE), and Cancer Therapeutics Response Portal (CTRP)], patient-derived tumor xenografts (PDX), and The Cancer Genome Atlas (TCGA). GDSC was used for training and cross-validation tests, and CCLE, CTRP, PDX, and TCGA were used for external validation. We performed ablation studies for the three stages and verified that the use of multi-omics data guarantees better performance of drug response prediction. Our results verified that Super.FELT outperformed the other methods at external validation on PDX and TCGA and was good at cross-validation on GDSC and external validation on CCLE and CTRP. In addition, through our experiments, we confirmed that using multi-omics data is useful for external non-cell line data.

CONCLUSION

By separating the three stages, Super.FELT achieved better performance than the other methods. Through our results, we found that it is important to train encoders and a classifier independently, especially for external test on PDX and TCGA. Moreover, although gene expression is the most powerful data on cell line data, multi-omics promises better performance for external validation on non-cell line data than gene expression data. Source codes of Super.FELT are available at  https://github.com/DMCB-GIST/Super.FELT .

Learning curves for drug response prediction in cancer cell lines

BACKGROUND

Motivated by the size and availability of cell line drug sensitivity data, researchers have been developing machine learning (ML) models for predicting drug response to advance cancer treatment. As drug sensitivity studies continue generating drug response data, a common question is whether the generalization performance of existing prediction models can be further improved with more training data.

METHODS

We utilize empirical learning curves for evaluating and comparing the data scaling properties of two neural networks (NNs) and two gradient boosting decision tree (GBDT) models trained on four cell line drug screening datasets. The learning curves are accurately fitted to a power law model, providing a framework for assessing the data scaling behavior of these models.

RESULTS

The curves demonstrate that no single model dominates in terms of prediction performance across all datasets and training sizes, thus suggesting that the actual shape of these curves depends on the unique pair of an ML model and a dataset. The multi-input NN (mNN), in which gene expressions of cancer cells and molecular drug descriptors are input into separate subnetworks, outperforms a single-input NN (sNN), where the cell and drug features are concatenated for the input layer. In contrast, a GBDT with hyperparameter tuning exhibits superior performance as compared with both NNs at the lower range of training set sizes for two of the tested datasets, whereas the mNN consistently performs better at the higher range of training sizes. Moreover, the trajectory of the curves suggests that increasing the sample size is expected to further improve prediction scores of both NNs. These observations demonstrate the benefit of using learning curves to evaluate prediction models, providing a broader perspective on the overall data scaling characteristics.

CONCLUSIONS

A fitted power law learning curve provides a forward-looking metric for analyzing prediction performance and can serve as a co-design tool to guide experimental biologists and computational scientists in the design of future experiments in prospective research studies.

PDXGEM: patient-derived tumor xenograft-based gene expression model for predicting clinical response to anticancer therapy in cancer patients

Background

Cancer is a highly heterogeneous disease with varying responses to anti-cancer drugs. Although several attempts have been made to predict the anti-cancer therapeutic responses, there remains a great need to develop highly accurate prediction models of response to the anti-cancer drugs for clinical applications toward a personalized medicine. Patient derived xenografts (PDXs) are preclinical cancer models in which the tissue or cells from a patient’s tumor are implanted into an immunodeficient or humanized mouse. In the present study, we develop a bioinformatics analysis pipeline to build a predictive gene expression model (GEM) for cancer patients’ drug responses based on gene expression and drug activity data from PDX models.

Results

Drug sensitivity biomarkers were identified by performing an association analysis between gene expression levels and post-treatment tumor volume changes in PDX models. We built a drug response prediction model (called PDXGEM) in a random-forest algorithm by using a subset of the drug sensitvity biomarkers with concordant co-expression patterns between the PDXs and pretreatment cancer patient tumors. We applied the PDXGEM to several cytotoxic chemotherapies as well as targeted therapy agents that are used to treat breast cancer, pancreatic cancer, colorectal cancer, or non-small cell lung cancer. Significantly accurate predictions of PDXGEM for pathological response or survival outcomes were observed in extensive independent validations on multiple cancer patient datasets obtained from retrospective observational studies and prospective clinical trials.

Conclusion

Our results demonstrated the strong potential of using molecular profiles and drug activity data of PDX tumors in developing a clinically translatable predictive cancer biomarkers for cancer patients. The PDXGEM web application is publicly available at http://pdxgem.moffitt.org.

A survey and systematic assessment of computational methods for drug response prediction

Drug response prediction arises from both basic and clinical research of personalized therapy, as well as drug discovery for cancers. With gene expression profiles and other omics data being available for over 1000 cancer cell lines and tissues, different machine learning approaches have been applied to drug response prediction. These methods appear in a body of literature and have been evaluated on different datasets with only one or two accuracy metrics. We systematically assess 17 representative methods for drug response prediction, which have been developed in the past 5 years, on four large public datasets in nine metrics. This study provides insights and lessons for future research into drug response prediction.

IgG Galactosylation status combined with MYOM2-rs2294066 precisely predicts anti-TNF response in ankylosing spondylitis

Background

Tumor necrosis factor (TNF) blockers have a high efficacy in treating Ankylosing Spondylitis (AS), yet up to 40% of AS patients show poor or even no response to this treatment. In this paper, we aim to build an approach to predict the response prior to clinical treatment.

Methods

AS patients during the active progression were included and treated with TNF blocker for 3 months. Patients who do not fulfill ASASAS40 were considered as poor responders. The Immunoglobulin G galactosylation (IgG-Gal) ratio representing the quantity of IgG galactosylation was calculated and candidate single nucleotide polymorphisms (SNPs) in patients treated with etanercept was obtained. Machine-learning models and cross-validation were conducted to predict responsiveness.

Results

Both IgG-Gal ratio at each time point and differential IgG-Gal ratios between week 0 and weeks 2, 4, 8, 12 showed significant difference between responders and poor-responders. Area under curve (AUC) of the IgG-Gal ratio prediction model was 0.8 after cross-validation, significantly higher than current clinical indexes (C-reactive protein (CRP) = 0.65, erythrocyte sedimentation rate (ESR) = 0.59). The SNP MYOM2-rs2294066 was found to be significantly associated with responsiveness of etanercept treatment. A three-stage approach consisting of baseline IgG-Gal ratio, differential IgG-Gal ratio in 2 weeks, and rs2294066 genotype demonstrated the ability to precisely predict the response of anti-TNF therapy (100% for poor-responders, 98% for responders).

Conclusions

Combination of different omics can more precisely to predict the response of TNF blocker and it is potential to be applied clinically in the future.

Electronic supplementary material

The online version of this article (10.1186/s10020-019-0093-2) contains supplementary material, which is available to authorized users.

Predicting drug response of tumors from integrated genomic profiles by deep neural networks

BACKGROUND

The study of high-throughput genomic profiles from a pharmacogenomics viewpoint has provided unprecedented insights into the oncogenic features modulating drug response. A recent study screened for the response of a thousand human cancer cell lines to a wide collection of anti-cancer drugs and illuminated the link between cellular genotypes and vulnerability. However, due to essential differences between cell lines and tumors, to date the translation into predicting drug response in tumors remains challenging. Recently, advances in deep learning have revolutionized bioinformatics and introduced new techniques to the integration of genomic data. Its application on pharmacogenomics may fill the gap between genomics and drug response and improve the prediction of drug response in tumors.

RESULTS

We proposed a deep learning model to predict drug response (DeepDR) based on mutation and expression profiles of a cancer cell or a tumor. The model contains three deep neural networks (DNNs), i) a mutation encoder pre-trained using a large pan-cancer dataset (The Cancer Genome Atlas; TCGA) to abstract core representations of high-dimension mutation data, ii) a pre-trained expression encoder, and iii) a drug response predictor network integrating the first two subnetworks. Given a pair of mutation and expression profiles, the model predicts IC50 values of 265 drugs. We trained and tested the model on a dataset of 622 cancer cell lines and achieved an overall prediction performance of mean squared error at 1.96 (log-scale IC50 values). The performance was superior in prediction error or stability than two classical methods (linear regression and support vector machine) and four analog DNN models of DeepDR, including DNNs built without TCGA pre-training, partly replaced by principal components, and built on individual types of input data. We then applied the model to predict drug response of 9059 tumors of 33 cancer types. Using per-cancer and pan-cancer settings, the model predicted both known, including EGFR inhibitors in non-small cell lung cancer and tamoxifen in ER+ breast cancer, and novel drug targets, such as vinorelbine for TTN-mutated tumors. The comprehensive analysis further revealed the molecular mechanisms underlying the resistance to a chemotherapeutic drug docetaxel in a pan-cancer setting and the anti-cancer potential of a novel agent, CX-5461, in treating gliomas and hematopoietic malignancies.

CONCLUSIONS

Here we present, as far as we know, the first DNN model to translate pharmacogenomics features identified from in vitro drug screening to predict the response of tumors. The results covered both well-studied and novel mechanisms of drug resistance and drug targets. Our model and findings improve the prediction of drug response and the identification of novel therapeutic options.

Clinical Drug Response Prediction by Using a Lq Penalized Network-Constrained Logistic Regression Method

BACKGROUND/AIMS

One of the most important impacts of personalized medicine is the connection between patients' genotypes and their drug responses. Despite a series of studies exploring this relationship, the predictive ability of such analyses still needs to be strengthened.

METHODS

Here we present the Lq penalized network-constrained logistic regression (Lq-NLR) method to meet this need, in which the predictors are integrated into the gene expression data and biological network knowledge and are combined with a more aggressive penalty function. Response prediction models for two cancer targeting drugs (erlotinib and sorafenib) were developed from gene expression data and IC50 values from a large panel of cancer cell lines by utilizing the proposed approach. Then the drug responders were tested with the baseline tumor gene expression data, yielding an in vivo drug sensitivity prediction.

RESULTS

These results demonstrated the high effectiveness of this approach. One of the best results achieved by our method was a correlation of 0.841 between the cell line in vitro drug response and patient's in vivo drug response. We then applied these two drug prediction models to develop a personalized medicine approach in which the subsequent treatment depends on each patient's gene-expression profile.

CONCLUSION

The proposed method is much better than the existing approach and can capture a more accurate reflection of the relationship between genotypes and phenotypes.

Machine learning and feature selection for drug response prediction in precision oncology applications

In-depth modeling of the complex interplay among multiple omics data measured from cancer cell lines or patient tumors is providing new opportunities toward identification of tailored therapies for individual cancer patients. Supervised machine learning algorithms are increasingly being applied to the omics profiles as they enable integrative analyses among the high-dimensional data sets, as well as personalized predictions of therapy responses using multi-omics panels of response-predictive biomarkers identified through feature selection and cross-validation. However, technical variability and frequent missingness in input "big data" require the application of dedicated data preprocessing pipelines that often lead to some loss of information and compressed view of the biological signal. We describe here the state-of-the-art machine learning methods for anti-cancer drug response modeling and prediction and give our perspective on further opportunities to make better use of high-dimensional multi-omics profiles along with knowledge about cancer pathways targeted by anti-cancer compounds when predicting their phenotypic responses.

Drug response prediction by ensemble learning and drug-induced gene expression signatures

Chemotherapeutic response of cancer cells to a given compound is one of the most fundamental information one requires to design anti-cancer drugs. Recently, considerable amount of drug-induced gene expression data has become publicly available, in addition to cytotoxicity databases. These large sets of data provided an opportunity to apply machine learning methods to predict drug activity. However, due to the complexity of cancer drug mechanisms, none of the existing methods is perfect. In this paper, we propose a novel ensemble learning method to predict drug response. In addition, we attempt to use the drug screen data together with two novel signatures produced from the drug-induced gene expression profiles of cancer cell lines. Finally, we evaluate predictions by in vitro experiments in addition to the tests on data sets. The predictions of the methods, the signatures and the software are available from http://mtan.etu.edu.tr/drug-response-prediction/.

Drug response prediction in high-risk multiple myeloma

A Drug Response Prediction (DRP) score was developed based on gene expression profiling (GEP) from cell lines and tumor samples. Twenty percent of high-risk patients by GEP70 treated in Total Therapy 2 and 3A have a progression-free survival (PFS) of more than 10years. We used available GEP data from high-risk patients by GEP70 at diagnosis from Total Therapy 2 and 3A to predict the response by the DRP score of drugs used in the treatment of myeloma patients. The DRP score stratified patients further. High-risk myeloma with a predicted sensitivity to melphalan by the DRP score had a prolonged PFS, HR=2.4 (1.2-4.9, P=0.014) and those with predicted sensitivity to bortezomib had a HR 5.7 (1.2-27, P=0.027). In case of predicted sensitivity to bortezomib, a better response to treatment was found (P=0.022). This method may provide us with a tool for identifying candidates for effective personalized medicine and spare potential non-responders from suffering toxicity.

Association analysis of the perturbation of interactions in biological pathways and anticancer drug activity

Understanding how different genomic mutational landscapes in patients with cancer lead to different responses to anticancer drugs is an important challenge for realizing precision medicine for cancer. Many studies have analyzed the comprehensive anticancer drug-response profiles and genomic profiles of cancer cell lines to identify the relationship between the anticancer drug response and genomic alternations. However, few studies have focused on interpreting these profiles with a network perspective. In this work, we analyzed genomic alterations in cancer cell lines by considering which interactions in the signaling pathway were perturbed by mutations. With our interaction-centric approach, we identified novel interaction/drug response associations for two drugs (afatinib and ixabepilone) for which no gene-centric association could be found. When we compared the performance of classifiers for predicting the responses to 164 drugs, the classifiers trained with interaction-centric features outperformed the classifiers trained with gene-centric features, despite the smaller number of features (p-value = 2.0 × 10(-3)). By incorporating the interaction information from signaling pathways, we revealed associations between genomic alterations and drug responses that could be missed when using a gene-centric approach.