Quantitative scoring of differential drug sensitivity for individually optimized anticancer therapies.

Yadav B, Pemovska T, Szwajda A, Kulesskiy E, Kontro M, Karjalainen R, Majumder MM, Malani D, Murumägi A, Knowles J, Porkka K, Heckman C, Kallioniemi O, Wennerberg K, Aittokallio T. · 2014

Read Quantitative scoring of differential drug sensitivity for individually optimized anticancer therapies. by Yadav B, Pemovska T, Szwajda A, Kulesskiy E, Kontro M, Karjalainen R, Majumder MM, Malani D, Murumägi A, Knowles J, Porkka K, Heckman C, Kallioniemi O, Wennerberg K, Aittokallio T. (2014) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/europmc-pmc4046135.txt.

Introduction

Bhagwan Yadav and colleagues build a scoring system that ranks drugs by how selectively they kill a patient's cancer cells, in Scientific Reports in 2014. Labs test hundreds of compounds on cells across many doses. Turning those curves into one fair score per drug is tricky, and noise can warp simple measures.

Their drug sensitivity score models the full dose curve and folds several traits into one number. It proved steadier than standard measures against technical noise. The sharper test came with leukaemia samples. Comparing patient cells with healthy controls picked out drugs that hit the cancer and spared normal cells. It also split patients into sensitive and resistant groups. Repeated testing tracked how responses shifted as disease advanced and returned. The team releases the tool as open source. The aim: turn lab screens into choices fitted to one patient.

Catalog record

CREATORYadav B, Pemovska T, Szwajda A, Kulesskiy E, Kontro M, Karjalainen R, Majumder MM, Malani D, Murumägi A, Knowles J, Porkka K, Heckman C, Kallioniemi O, Wennerberg K, Aittokallio T.
YEAR2014
COLLECTIONPAPERS
FORMATTXT
FILE SIZE40 KB
LICENSECC BY 4.0
RIGHTSCC BY
SOURCEEurope PMC

Source: https://europepmc.org/articles/PMC4046135

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Quantitative scoring of differential drug sensitivity for individually optimized anticancer therapies By Bhagwan Yadav, Tea Pemovska, Agnieszka Szwajda, Evgeny Kulesskiy, Mika Kontro, Riikka Karjalainen, Muntasir Mamun Majumder, Disha Malani, Astrid Murumägi, Jonathan Knowles, Kimmo Porkka, Caroline Heckman Published in: Scientific Reports DOI: 10.1038/srep05193 ABSTRACT We developed a systematic algorithmic solution for quantitative drug sensitivity scoring (DSS), based on continuous modeling and integration of multiple dose-response relationships in high-throughput compound testing studies. Mathematical model estimation and continuous interpolation makes the scoring approach robust against sources of technical variability and widely applicable to various experimental settings, both in cancer cell line models and primary patient-derived cells. Here, we demonstrate its improved performance over other response parameters especially in a leukemia patient case study, where differential DSS between patient and control cells enabled identification of both cancer-selective drugs and drug-sensitive patient sub-groups, as well as dynamic monitoring of the response patterns and oncogenic driver signals during cancer progression and relapse in individual patient cells ex vivo. An open-source and easily extendable implementation of the DSS calculation is made freely available to support its tailored application to translating drug sensitivity testing results into clinically actionable treatment options. Cell-based compound screening provides a rich functional readout for many biomedical applications. In cancer research, the possibility to profile cellular responses to an extensive collection of anti-cancer compounds enables a systematic means to repurpose existing drugs to new indications, identify druggable vulnerabilities in various types of cancer cells and to functionally investigate cellular pathways behind drug sensitivity or resistance. Recent studies have successfully explained or even predicted drug responses by means of genetic aberrations or other genomic biomarkers in wide panels of cancer cell lines1234. Similar large-scale drug testing efforts in primary cancer samples are increasingly being carried out to enable functional investigation of cellular addictions in individual cancer patients; for instance, to predict pathway dependencies and to identify potential therapeutic options for leukemia patients56. Systematic profiling of the relative activity of hundreds or thousands of drugs at several concentrations in a large number of cancer samples or cell types facilitates the stratification of cancer patients and disease subtypes, as well as development of personalized treatment strategies for clinical applications. However, high-throughput drug testing experiments often result in high-dimensional sample-dose-response matrices, with inherent measurement noise and technical variability, which hinders many downstream analyses, such as those aimed at detecting differential drug sensitivities or clustering of patients and/or drugs based on their selective response patterns. To provide quantitative information about the degree of drug efficacy in a given sample, the dose dimension of these matrices is often summarized into single response parameter estimated form dose-response models, such as IC50 or EC50 (half-maximal inhibitory/effective concentration)123578. Although proven sufficient in many applications, any single model parameter can capture only limited information about the differences in the response patterns9, especially when comparing cancer and normal cells. Recently, an ‘Activity Area' metric was used to estimate both the efficacy and potency of 24 compounds in hundreds of cancer cell lines3. This type of discrete approximation, based on summing up the observed responses at each dose level, was shown to perform well under controlled in vitro settings with relatively densely-sampled concentration ranges and narrow bioactivity spectra3. Here, we developed and implemented a quantitative scoring approach, named drug sensitivity score (DSS), which captures and integrates the multiparametric dose-response relationships into a single metric to identify selective drug response patterns between cancer and control cells, rather than scoring drug activity in cancer cells alone. Analytic integration of the area under the non-linear dose-response model combines the advantages of both the model-based and area-based response calculations. Applications of DSS to drug sensitivity testing of acute myeloid leukemia (AML) patient cells ex vivo demonstrated its improved performance, also when profiling larger compound panels and broader bioactivity spectra at sparsely-sampled dose levels (10,000-fold range) in fresh, primary cells. Several case studies in in vitro models from the Cancer Cell Line Encyclopedia (CCLE) resource3 also supported the applicability of the DSS metric to various experimental settings and application cases, where the aim is to identify both sensitive and selective drug response patterns. To promote its application to the future drug testing studies, we have made publicly available an open-source and easily extendable implementation of the model-based DSS calculations in the form of a stand-alone R-package. RESULTS Our quantitative scoring approach is based on closed-form integration of the area under the estimated dose-response curve (AUC; Figure 1a); the generic modeling approach can be used in the context of standard logistic, sigmoidal or Hill slope response functions (Figure 1b). The continuous model estimation and interpolation effectively summarize the complex dose-response relationship into a single response metric, named DSS (Supplementary Fig. 1b). More formally, if R(x) models the normalized drug response at a concentration x, then the integral response I over the dose range that exceeds a given minimum activity level, Amin, is calculated analytically as a continuous function of multiple parameters of the non-linear response model, including its slope at IC50 as well as the top and bottom asymptotes of the response (Rmax and Rmin).: Importantly, differential DSS (dDSS) quantifies the selective response of cancer cells, relative to that of control cells, when control samples are available; dDSS is calculated by the difference between drug response quantified in patient cells (patient DSS) and the average drug response of control samples (controls DSS) (Figure 1c). To discriminate those compounds which are effective at higher concentrations only (potential toxic off-target responses), and to favor those that show potency over a relative wide therapeutic window, the analytic AUC calculation (referred to as DSS1) was further normalized by the logarithm of the top asymptote Rmax(DSS2) and by the dose range over which the response exceeds the activity threshold Amin (DSS3), respectively (mathematical derivation of the closed-form solutions when using four-parameter logistic response model is given in Supplementary Methods). The DSS R-package and its source code are freely available at https://dss-calculation.googlecode.com/svn/trunk/. DSS calculation improves drug response profiling in primary leukemic cells We initially developed and implemented the DSS calculation in the context of our ongoing drug sensitivity and resistance testing (DSRT) program, with the aim to provide informed choices for clinicians on the treatment of relapsed or chemorefractory acute myeloid leukemia (AML) patients based on the ex vivo DSRT results of the patient cells6. The screening panel of 204 compounds used in this study covers virtually all FDA-approved small molecule anti-cancer drugs, along with a collection of emerging, investigational and preclinical oncology compounds, including signal transduction inhibitors targeting major oncogenic signaling pathways (Supplementary Table 1). The drugs were plated at 5 concentrations in 10-fold dilution series. The challenge here was to score the individual drug sensitivities in a patient testing setup, where limited sources of fresh, primary cancer cells are available to quantify selective responses in comparison to control cells from healthy donors. Here, we functionally profiled 22 bone marrow aspirates from 14 AML patients, whereas 4 bone marrow samples from healthy donors tested in the same way were used as controls. In the present study, a total of 5,161 sample-compound pairs were analyzed using the DSS analysis pipeline (Figure 1). To test its quantitative performance, we first systematically evaluated the predictive power of DSS in

Read the complete text