US-20260128148-A1 - Classifying Images of Dose-Response Graphs

US20260128148A1US 20260128148 A1US20260128148 A1US 20260128148A1US-20260128148-A1

Abstract

A computer-implemented method of classifying images comprising dose-response graphs obtained from dose-response experiments. The method comprises receiving, at a curve shape classifier model, an input comprising image data including a plurality of pixels, wherein the image data represents an image of a dose-response graph indicating a relationship between the concentration of a compound and its activity. The curve shape classifier model comprises a neural network model configured for classifying images of dose-response graphs into a plurality of dose-response graph categories relating to curve shape. The method further comprises generating, using the neural network model, a classification output for the image represented by the received image data, said generating comprising processing the image data using one or more layers of the neural network model in accordance with parameters associated with the one or more layers.

Inventors

Marc BIANCIOTTO
Kun Mi

Assignees

SANOFI

Dates

Publication Date: 20260507
Application Date: 20260105
Priority Date: 20201130

Claims (14)

1 . A computer-implemented method of classifying images comprising dose-response graphs obtained from dose-response experiments, comprising: receiving, at a curve shape classifier model, an input comprising image data including a plurality of pixels, wherein the image data represents an image of a dose-response graph indicating a relationship between a concentration of a compound and its activity, wherein the curve shape classifier model comprises a neural network model configured for classifying images of dose-response graphs into a plurality of shape categories, wherein each shape category is associated with a respective curve shape which defines a dose-response relationship for the category; and generating, using the neural network model, a classification output for the image represented by the received image data, the generating comprising processing the image data using one or more layers of the neural network model in accordance with parameters associated with the one or more layers.
2 . The computer-implemented method of claim 1 , wherein the curve shape classifier model comprises a convolutional neural network model.
3 . The computer-implemented method of claim 1 , comprising classifying a dose-response graph into a first or second dispersion category based on differences between measures of activity at the same concentration, wherein image data representing an image of the dose-response graph is processed using the curve shape classifier model if the dose-response graph is classified in the first dispersion category.
4 . The computer-implemented method of claim 3 , wherein the dose-response graph is classified into a first or second dispersion category based on quartile values over a difference in measures of activity as a function of concentration.
5 . The computer-implemented method of claim 3 , comprising classifying the dose-response graph into a first or second dispersion category using a dispersion classifier comprising a multi-layer perceptron neural network model.
6 . The computer-implemented method of claim 1 , wherein the plurality of shape categories includes one or more of: a category for high activity across the whole concentration range; a category for sigmoid curves in which an upper asymptotic part is visible but a lower asymptotic part is not; a category for well behaved sigmoid curves which include lower and upper asymptotic parts; a category for sigmoid curves in which the lower asymptotic part is visible but an upper asymptotic part is not, and which reaches a 50% activity threshold, wherein part of the dose-response graph after an inflexion point is visible; a category for weakly active compounds in the concentration range of the dose-response graph; a category for non-active compounds in the concentration range of the dose-response graph; a high slope category for sigmoid curves with a high slope at the EC50; a low slope category for sigmoid curves with a low slope at the EC50; a category for sigmoid curves in which the the difference between A(c) at upper and lower asymptotes is less than 70%; a category in which there is an alternative increase and decrease of activity with respect to concentration; and a category in which no activity is shown except for the highest concentration or two highest concentrations.
7 . The computer-implemented method of claim 1 , further comprising performing pre-processing including receiving raw data representing a set of data points for the dose-response graph and generating the image of the dose-response graph based on the raw data.
8 . The computer-implemented method of claim 7 , comprising: receiving raw data representing a set of data points for each of a plurality of dose-response graphs; generating a respective image for each dose-response graph, wherein the image comprises a plurality of pixels and depicts at least some of the respective set of data points relative to Cartesian axes, wherein each image is generated with the same pixel height and pixel width, wherein the Cartesian axes are positioned at the same location in each image; and receiving, at the curve shape classifier model, image data for each respective image.
9 . The computer-implemented method of claim 8 , wherein each image has a vertical axis with the same scale.
10 . A computer-implemented method of producing a curve shape classifier model for classifying dose-response graphs obtained from dose-response experiments, comprising: receiving a plurality of training images at a neural network model, wherein each training image is an image of a dose-response graph indicating a relationship between a concentration of a compound and its activity; generating an output for each training image, wherein generating the output for a training image comprises processing the training image through one or more layers of the neural network model in accordance with parameters associated with the one or more layers; and updating the parameters based on an objective function comprising a comparison between the generated output for each training image with corresponding label data associated with the training image, the label data indicating that the training image belongs to one or more shape categories, wherein each shape category is associated with a respective curve shape which defines a dose-response relationship for the category.
11 . The computer-implemented method of claim 10 , wherein each training image depicts a respective set of data points relative to Cartesian axes, wherein each training image has the same pixel height and pixel width, wherein the Cartesian axes are positioned at the same location in each image.
12 . The computer-implemented method of claim 10 , wherein the one or more shape categories include: a bell-shaped curve category; and a toxicity category.
13 . A data processing apparatus comprising: one or more processors configured to perform a computer-implemented method of classifying images comprising dose-response graphs obtained from dose-response experiments, the method comprising: receiving, at a curve shape classifier model, an input comprising image data including a plurality of pixels, wherein the image data represents an image of a dose-response graph indicating a relationship between a concentration of a compound and its activity, wherein the curve shape classifier model comprises a neural network model configured for classifying images of dose-response graphs into a plurality of shape categories, wherein each shape category is associated with a respective curve shape which defines a dose-response relationship for the category; and generating, using the neural network model, a classification output for the image represented by the received image data, the generating comprising processing the image data using one or more layers of the neural network model in accordance with parameters associated with the one or more layers.
14 . A non-transitory computer-readable storage medium comprising instructions, which when executed by one or more processors, cause the one or more processors to perform a computer-implemented method of classifying images comprising dose-response graphs obtained from dose-response experiments, the method comprising: receiving, at a curve shape classifier model, an input comprising image data including a plurality of pixels, wherein the image data represents an image of a dose-response graph indicating a relationship between a concentration of a compound and its activity, wherein the curve shape classifier model comprises a neural network model configured for classifying images of dose-response graphs into a plurality of shape categories, wherein each shape category is associated with a respective curve shape which defines a dose-response relationship for the category; and generating, using the neural network model, a classification output for the image represented by the received image data, the generating comprising processing the image data using one or more layers of the neural network model in accordance with parameters associated with the one or more layers.

Description

CROSS REFERENCE TO RELATED APPLICATIONS The present application is a continuation of U.S. patent application Ser. No. 18/038,300, filed on May 23, 2023, which is the national stage entry of International Patent Application No. PCT/EP 2021/083404, filed on Nov. 29, 2021, and claims priority to Application No. EP 20315469.5, filed on Nov. 30, 2020, the disclosures of which are incorporated herein by reference. TECHNICAL FIELD The present disclosure relates to a computer-implemented method of classifying images comprising dose-response graphs obtained from a plurality of dose-response experiments. BACKGROUND At the dawn of drug discovery projects, a potential drug target is identified. The drug target is a molecule in the body, that is intrinsically associated with a particular disease process. Depending upon the disease to be treated, the target may be a protein (e.g. a receptor protein, an enzyme, an ion channel or a transporter protein) or a nucleic acid (e.g. DNA). Researchers hypothesise that modifying the activity of the target with a drug will result in a desirable therapeutic effect. Once a target has been identified, a test system is identified or developed which produces a detectable signal to assess the effect of compounds on the drug target. This test system is called an assay. Once an assay has been identified or developed, researchers can use it to identify compounds which have the desired activity. Typically, a compound will be tested at a number of concentrations and a dose-response (DR) graph can be generated. Analysis of the DR graph allows researchers to determine if a compound is active, and at what concentration. Where it is desirable to test a large number of potential compounds High Throughput Screening (HTS) is often used. This uses robotics, data processing/control software, liquid handling devices and sensitive detectors, and allows researchers to quickly conduct thousands or even millions of screening tests. When testing all compounds at a number of concentrations to generate DR data, this is called quantitative HTS (qHTS). However, the large amount of data generated at the dose-response (DR) step of a HTS campaign requires careful analysis by the researchers in order to detect artifacts and correct erroneous data points before validating the experiments. This step, which requires expert review of each DR experiment can be time consuming and prone to human errors or inconsistencies. SUMMARY This specification describes a computer-implemented method of classifying images comprising dose-response graphs obtained from dose-response experiments. The method comprises receiving, at a curve shape classifier model, an input comprising image data including a plurality of pixels, wherein the image data represents an image of a dose-response graph indicating a relationship between the concentration of a compound and its activity. The curve shape classifier model comprises a neural network model configured for classifying images of dose-response graphs into a plurality of dose-response graph categories relating to curve shape. In particular, each dose-response graph category may be associated with a respective curve shape which defines the dose-response relationship for the category. The method includes generating, using the neural network model, a classification output for the image represented by the received image data, said generating comprising processing the image data using one or more layers of the neural network model in accordance with parameters associated with the one or more layers. The neural network model may comprise a convolutional neural network model. Classifying images of dose-response graphs (rather than raw dose-response graph data) reduces the impact of any lack of homogeneity between different inputs to the classifier model, or between inputs used for prediction and the training data. Thus, the claimed approach provides for a flexible classifier which can classify dose-response graph images with e.g. different number of data points (e.g. 8, 10 or 12 concentrations), missing points and/or with different numbers of replicates. The classification output may comprise a vector of probabilities, each probability representing a likelihood that the input represented by the received image data belongs to a respective one of the categories. Alternatively, or in addition, the classification output may indicate the category having the highest probability. In some examples the classification output may indicate the two (or more) most probable categories. In an example implementation, a dose-response graph may be initially classified into a first or second dispersion category, wherein image data representing an image of the dose-response graph is processed using the curve shape classifier model only if the dose-response graph is classified in the first dispersion category. A dispersion classifier may be used to classify the dose-response graph into the first or second dispersion catego