# US 2021/0241911 A1 · Method for extracting intrinsic properties of cancer cells from gene expression profiles of cancer patients and device for the same

> Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt

- Number: US 2021/0241911 A1
- Document ID: US20210241911A1
- Status: Published application
- Kind: A1
- Application number: 17/163,459
- Filed: 2021-01-31
- Published: 2021-08-05
- Earliest claimed: 2020-01-31
- Assignees: Korea Advanced Institute of Science and Technology
- Applicants: Korea Advanced Institute of Science and Technology
- Inventors: Kwang-Hyun CHO; Dongsan KIM
- Web page: https://patentgrep.com/patent/US20210241911A1

## CPC

- G16H 50/20: for computer-aided diagnosis, e.g. based on medical expert systems
- G06F 18/2133: based on naturality criteria, e.g. with non-negative factorisation or negative correlation
- G16B 25/10: Gene or protein expression profiling; Expression-ratio estimation or normalisation
- G16B 40/20: Supervised data analysis
- G16H 10/40: for data related to laboratory analysis, e.g. patient specimen analysis
- G16H 50/70: for mining of medical data, e.g. analysing previous cases of other patients
- G16H: HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
- G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N)
- G16B: BIOINFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR GENETIC OR PROTEIN-RELATED DATA PROCESSING IN COMPUTATIONAL MOLECULAR BIOLOGY

## Abstract

Provided is a method of predicting a condition of a patient, including decomposing an input matrix to produce a residual matrix, the input matrix representing patients and expression levels of genes of the patients, training a classifier by using health condition values the patients as learning criteria and inputting the residual matrix into the classifier, and obtaining a value for predicting a health condition of a first patient, by inputting a first input matrix including the expression levels of genes of the first patient into the classifier, the first residual matrix being a residual matrix obtained from the first input matrix by using the predetermined algorithm.

## Claims

### Claim 1

A method of predicting a condition of a patient, comprising:
analyzing, by a gene expression level analysis device, expression levels of genes of a first patient of which a health condition is required to be predicted;
generating, by a computing device, a first input matrix including the expression levels of genes of the first patient based on the expression levels of genes analyzed for the first patient by receiving the expression levels of genes analyzed for the first patient from the gene expression level analysis device; and
obtaining, by the computing device, a value for predicting the health condition of the first patient by inputting the first input matrix into an input layer of a classifier or inputting a first residual matrix into the input layer of the classifier, the classifier being trained by a predetermined supervised learning method, the first residual matrix being a residual matrix obtained from the first input matrix by using a predetermined algorithm,
wherein the predetermined algorithm is an algorithm for decomposing a matrix input to the algorithm into a sum of a product of a plurality of element matrices and one residual matrix, and
the predetermined supervised learning method comprises:
analyzing, by the gene expression level analysis device, expression levels of genes from respective cells obtained from a plurality of patients;
calculating, by the computing device, an input matrix representing the plurality of patients and expression levels of a plurality of genes of each of the patients based on the expression levels of genes analyzed for the plurality of patients, by receiving the expression levels of genes analyzed for the plurality of patients from the gene expression level analysis device;
decomposing, by the computing device, the input matrix into a sum of a product of a plurality of element matrices and one residual matrix by using the algorithm; and
training, by the computing device, the classifier by supervised learning by using a plurality of health condition values representing health conditions of the plurality of patients as learning criteria and inputting the residual matrix into the input layer of the classifier, in order to train the classifier receiving a matrix having the same dimension as the residual matrix by the supervised learning.

### Claim 2

A method of predicting a condition of a patient, comprising:
decomposing, by a computing device, an input matrix into a sum of a product of a plurality of element matrices and one residual matrix by using a predetermined algorithm, the input matrix representing expression levels of a plurality of genes of a plurality of patients and each of the patients;
training, by the computing device, a classifier by supervised learning by using a plurality of health condition values representing health conditions of the plurality of patients as learning criteria and inputting the residual matrix into the input layer of the classifier, in order to train the classifier receiving a matrix having the same dimension as the residual matrix by the supervised learning; and
obtaining, by the computing device, a value for predicting a health condition of a first patient, by inputting a first input matrix into an input layer of the classifier or inputting a first residual matrix into the input layer of the classifier, the first input matrix including the expression levels of genes of the first patient of which a health condition is required to be predicted, the first residual matrix being a residual matrix obtained from the first input matrix by using the algorithm,
wherein the predetermined algorithm is an algorithm for decomposing a matrix input to the algorithm into a sum of a product of a plurality of element matrices and one residual matrix.

### Claim 3 (depends on 2)

The method of claim 2, further comprising:
analyzing, by the gene expression level analysis device, expression levels of genes of a first patient of which the health condition is required to be predicted;
generating, by the computing device, the first input matrix including the expression levels of genes of the first patient based on the expression levels of genes analyzed for the first patient by receiving the expression levels of genes analyzed for the first patient from the gene expression level analysis device; and
obtaining, by the computing device, the value for predicting the health condition of the first patient, by inputting the first input matrix into an input layer of the classifier or inputting a first residual matrix into the input layer of the classifier, the first residual matrix being a residual matrix obtained from the first input matrix by using a predetermined algorithm.

### Claim 4 (depends on 2)

The method of claim 2, wherein the predetermined algorithm is a non-negative matrix factorization algorithm.

### Claim 5 (depends on 2)

The method of claim 2, wherein the classifier is trained by the supervised learning to classify the plurality of patients by receiving the residual matrix and assigning indexes representing different patients included in the input residual matrix to predetermined groups.

### Claim 6 (depends on 2)

The method of claim 2, wherein each row of the input matrix and the residual matrix represents a different gene, each column of the input matrix represents a different patient.

### Claim 7 (depends on 2)

The method of claim 2, wherein each row of the input matrix represents a different gene, and each column of the input matrix represents a different test sample collected from bodies of the patients, and
when patients represented by two different columns of the input matrix are the same patient, the two different columns are test samples collected from different sites of a body of the same patient.

### Claim 8 (depends on 2)

The method of claim 2, wherein the plurality of element matrices comprises a total of two element matrices, and
the input matrix is the same as a matrix obtained by adding the residual matrix to a product of the two element matrices.

### Claim 9

A system for predicting a health condition of a patient, comprising:
a gene expression level analysis device; and
a computing device,
wherein the gene expression level analysis device is configured to analyze expression levels of genes of a first patient of which a health condition is required to be predicted, and analyze expression levels of genes from respective cells obtained from a plurality of patients,
the computing device comprises:
a residual generation unit configured to output one residual matrix by receiving an input matrix representing the plurality of patients and expression levels of a plurality of genes of each of the patients and decomposing the input matrix into a sum of a product of a plurality of element matrices and the residual matrix; and
a classifier configured to receive a matrix having the same dimension as the residual matrix, the classifier being trained by a predetermined supervised learning method; and
a processing unit,
the processing unit is configured to:
generate a first input matrix including the expression levels of genes of the first patient based on the expression levels of genes analyzed for the first patient, by receiving the expression levels of genes analyzed for the first patient from the gene expression level analysis device; and
calculate a value for predicting the health condition of the first patient by inputting the generated first input matrix into an input layer of the classifier, and
the predetermined supervised learning method comprises:
analyzing, by the gene expression level analysis device, expression levels of genes from respective cells obtained from a plurality of patients;
calculating, by the computing device, an input matrix representing the plurality of patients and expression levels of a plurality of genes of each of the patients based on the expression levels of genes analyzed for the plurality of patients, by receiving the expression levels of genes analyzed for the plurality of patients from the gene expression level analysis device;
decomposing, by the residual generation unit of the computing device, the input matrix into a sum of a product of a plurality of element matrices and one residual matrix by using a predetermined algorithm; and
training, by the computing device, the classifier by supervised learning by using a plurality of health condition values representing health conditions of the plurality of patients as learning criteria and inputting the residual matrix into the input layer of the classifier, in order to train the classifier by the supervised learning.

### Claim 10 (depends on 9)

The system of claim 9, wherein the computing device further comprises a user interface, and
sizes of the plurality of element matrices are adjusted according to a parameter value input through the user interface.

### Claim 11 (depends on 9)

The system of claim 9, wherein the residual generation unit comprises an equator, and
the equator is configured to decompose the input matrix by using a non-negative matrix factorization (NMF) algorithm.

### Claim 12 (depends on 3)

The method of claim 3, wherein the predetermined algorithm is a non-negative matrix factorization algorithm.

### Claim 13 (depends on 3)

The method of claim 3, wherein the classifier is trained by the supervised learning to classify the plurality of patients by receiving the residual matrix and assigning indexes representing different patients included in the input residual matrix to predetermined groups.

### Claim 14 (depends on 3)

The method of claim 3, wherein each row of the input matrix and the residual matrix represents a different gene, each column of the input matrix represents a different patient.

### Claim 15 (depends on 3)

The method of claim 3, wherein each row of the input matrix represents a different gene, and each column of the input matrix represents a different test sample collected from bodies of the patients, and
when patients represented by two different columns of the input matrix are the same patient, the two different columns are test samples collected from different sites of a body of the same patient.

### Claim 16 (depends on 3)

The method of claim 3, wherein the plurality of element matrices comprises a total of two element matrices, and
the input matrix is the same as a matrix obtained by adding the residual matrix to a product of the two element matrices.