Search

EP-3520004-B1 - COMPUTER DEVICE FOR DETECTING AN OPTIMAL CANDIDATE COMPOUND AND METHODS THEREOF

EP3520004B1EP 3520004 B1EP3520004 B1EP 3520004B1EP-3520004-B1

Inventors

  • FUENTES, Emmanuel, Israel
  • AVINASH, GOPAL, BILIGERI
  • GRAVES, Robert, John
  • THATTE, Abhijit, Vijay
  • KODESH, Afek
  • CARON, JEFFERY
  • DAS, SHARMISTHA

Dates

Publication Date
20260506
Application Date
20171002

Claims (7)

  1. A computer-implemented method for detecting an optimal candidate compound based on image data of a plurality of samples comprising a cell line and one or more biomarkers, wherein the plurality of samples is arranged on one or more well-plates according to a plate map configuration, wherein the plate map configuration is providing locations of the plurality of samples arranged on the one or more well-plates, wherein the samples are exposed to the one or more biomarkers and different concentrations of a candidate compound forming at least one concentration gradient, the candidate compound being comprised in a plurality of candidate compounds, wherein the plate map configuration indicates the location of a candidate compound of the plurality of candidate compounds and a respective concentration of the candidate compound on the one or more well-plates, said method comprising: generating (310) phenotypic profiles of each concentration gradient of each of the plurality of candidate compounds at a plurality of successive points in time to form a plurality of compound profiles, wherein generating phenotypic profiles comprises the steps of obtaining (312) image data depicting each sample comprised in the concentration gradient, wherein the image data is depicting each sample comprised in the concentration gradient from a plurality of field of views, and generating (314) a class-label and a class for each cell of the samples based on the image data, wherein obtaining image data comprises retrieving and/or receiving the image data from a memory (115), an external node, an internal and/or external database or an image generator (220), selecting (315) an exemplary subset of image data depicting at least one cell of the samples for each class-label and/or class, displaying (316) the exemplary subset of image data and the respective class-label and/or class to a user, receiving (317) user input data from the user indicative of an operation on at least one class, performing (318) the operation on the class-label and/or class of cells of the samples based on the user input data, wherein the operation on at least one class is selected from add class, delete class, split class or merge class, wherein the phenotypic profiles of a concentration gradient comprise a cell count of each class for each sample comprised in the concentration gradient, where each sample comprises different concentrations of the respective candidate compound, and wherein a compound profile indicates a respective cell count over different concentrations and over time, obtaining (322) one or more reference compound profiles, calculating (324) a multi-dimensional differential value for each compound profile of the plurality of compound profiles based on the one or more reference compound profiles, and detecting (320) the optimal candidate compound by evaluating a comparison criterion on the plurality of compound profiles, wherein the comparison criterion is evaluated based on the multi-dimensional differential values.
  2. The method according to claim 1, wherein the image data is depicting each sample comprised in the concentration gradient processed with a plurality of image filters.
  3. A computer device (100) for detecting an optimal candidate compound based on a plurality of samples, the computer device comprising: a processor (112), and a memory (115), said memory (115) containing instructions executable by said processor, whereby said computer device is operative to perform the method of any of the preceding claims.
  4. The computer device (100) of claim 3, wherein the computer device (100) is further operative to display the phenotypic profiles and/or compound profiles on a display (118) of the computer device (100).
  5. The computer device (100) according to claim 3 or 4, wherein the concentration gradients of a candidate compound comprise a plurality of separate wells, wherein each well comprises a sample of the cell line exposed the one or more biomarkers and different concentrations of the candidate compound and is arranged according to the plate map configuration.
  6. A computer program comprising computer-executable instructions for causing a computer device (100), when the computer-executable instructions are executed on a processing unit comprised in the computer device (100), to perform the method of claim 1 or 2.
  7. A computer program product comprising a computer-readable storage medium, the computer-readable storage medium having the computer program according to claim 6 embodied therein.

Description

Technical Field The invention relates to a computer device for detecting an optimal candidate compound. Furthermore, the invention also relates to a corresponding method, a computer device, a computer program, and a computer program product. Background Testing and evaluating new compounds in experiments, e.g. compounds intended for medical use, may typically involve studying cellular mechanisms in a biological context to detect and/or select a candidate compound from a plurality of candidate compounds. Experiments comprising a plurality of samples, e.g. arranged according to a plate map configuration on determined locations a well-plate, may be conducted purely to discover or describe new insights into a biological phenotype. E.g. the observable characteristics or traits of cells under study. To interpret large, high dimensional (dense) cytometric data sets obtained from the experiments poses a problem for researchers and scientists, e.g. by generating statistical experiment insights and performing cell population classification. In particular, a disadvantage of conventional tools or systems for automatic classification of biological objects, such as cells in a sample, is that they require the user to leverage univariate and bivariate visualizations to determine cutoff regions and/or decision regions between classes and/or cell population regions. Further, determining decision regions between cell populations is rarely a linear clear-cut, population granularity may be difficult to understand, when analyzing cytometric data sets of an experiment visually by using a single measure and potentially performing iterative analysis of a single field-of-view (FOV). Further manual cytometric cellular measure investigations are labor intensive and it is difficult to select which cytometric cellular measure measures to us in the analysis and which contain the best ability to split a population and is therefore vulnerable to human error. Some conventional systems apply a two phase manual process involving visual cellular labeling and supervised cellular classification. The use of these systems involves the user to bin cellar objects as exemplars of a respective cell population manually by clicking cells and annotating them via manual text entry. A disadvantage of this is that it is a relatively slow and labor intensive process. As it is usually performed by viewing image data depicting single field of view, which rarely represents the entire cell population when a dose response is present, it has the disadvantage of reducing predictive modeling possibilities and thus limiting the potential inherent in the large and dense data sets available. A trend in conventional systems is to try and circumvent features extraction and/or cytometric features, as they are bound by human understanding, and instead rely on machine learning features, e.g. measures and/or features that may be extracted by a computer with less effort than cytometric features. Disadvantage of using machine learning features is that they are hard or impossible to understand or interpret by a user and that they therefore do not allow the user to evaluate and exclude the viability and/or deviation of input and/or intermediate data. "Multiplexed mass cytometry profiling of cellular states perturbed by small-molecule regulators" (Bodenmiller B. et al., Nature Biotechnology, pages 858-867, August 19, 2012) discloses mass-tag cellular barcoding (MCB), which increases mass cytometry throughput by using n metal ion tags to multiplex up to 2n samples. "Multi-parameter phenotypic profiling: using cellular effects to characterize small-molecule compounds" (Feng Y. et al., Nature Reviews Drug Discovery, pages 567-578, July 1, 2009) suggests a multi-parameter cellular phenotypic profiling as a promising solution. In this approach, broad and quantitative molecular and physiological measurements of cellular responses to compound treatment are used to provide information on compound activity and target mechanisms. "Disease Modeling and Phenotypic Drug Screening for Diabetic Cardiomyopathy using Human Induced Pluripotent Stem Cells " (Drawnel F. M. et al., Cell Reports, vol. 9, no. 3, pages 810-820, 2014) develops environmentally and genetically driven in vitro models of the condition using human-induced-pluripotent-stem-cell-derived cardiomyocytes. GB 2 434 225 A suggests a method of generating classification models to predict biological activity of a population of cells. US 2006/045348 A1 discloses an automated system that conducts, monitors, and validates a cell-line-based biological experiment including one or more treatment compounds or other external stimuli. "Review: imaging technologies for flow cytometry" (Yuanyuan Han et al., Lab on a Chip, vol. 16, no. 24, 01.01.2016, pages 4639-4647) relates to imaging flow cytometry technologies and their challenges. US 2009/0170091 A1 discloses a method for predicting biological responses resulting from exposure to the test subs