# US 2020/0402660 A1 · Diagnostic Process for Disease Detection using Gene Expression based Multi Layer PCA Classifier

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

- Number: US 2020/0402660 A1
- Document ID: US20200402660A1
- Status: Published application
- Kind: A1
- Application number: 16/888,722
- Filed: 2020-05-30
- Published: 2020-12-24
- Earliest claimed: 2019-06-20
- Inventors: Latha Chakravarthy; Prathik Abhay Chakravarthy; Sudarshan Venkat Chakravarthy; Vasu Devan Chakravarthy
- Web page: https://patentgrep.com/patent/US20200402660A1

## CPC

- G16H 50/20: for computer-aided diagnosis, e.g. based on medical expert systems
- G16B 25/10: Gene or protein expression profiling; Expression-ratio estimation or normalisation
- G16B 40/30: Unsupervised data analysis
- G16H 10/40: for data related to laboratory analysis, e.g. patient specimen analysis
- G16H 10/60: for patient-specific data, e.g. for electronic patient records
- G16H 50/30: for calculating health indices; for individual health risk assessment
- G16H 50/50: for simulation or modelling of medical disorders
- G16H 50/70: for mining of medical data, e.g. analysing previous cases of other patients
- G16H 70/60: relating to pathologies
- G16H: HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
- G16B: BIOINFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR GENETIC OR PROTEIN-RELATED DATA PROCESSING IN COMPUTATIONAL MOLECULAR BIOLOGY

## Abstract

A diagnostic process for disease detection using gene expression based PCA (Principal Component Analysis) classifier is provided. The present invention includes a method of diagnosing disease in a patient by performing a multilayer PCA classification that analyzes gene expression profiles of patients' biological samples to predict their class as disease or healthy, uses patients' biological samples to extract a set of fingerprint genes for specific disease and cell type which can be used as identification features, and classifies patient biological samples as disease or healthy based on differential gene expression profiles of the fingerprint genes. The present invention also implements the multilayer PCA classifier on patients' biological samples to extract a set of fingerprint genes for specific disease and cell type, and tabulates the fingerprint gene information in a database that will be referenced for disease diagnosis, prescreening in early stages of disease, and for confirming the stage of the disease, executed through the many embodiments of the invention.

## Claims

### Claim 1

A method of diagnosing disease in a patient comprising:
a multilayer Principal Component Analysis (PCA) Classifier that analyzes gene expression profile data of patients' biological samples and predicts their class as disease or healthy;
the system wherein is a classifier that analyzes gene expression profile data for specific disease and cell/tissue type to extract a set of fingerprint genes used as identification features for the specific disease and cell/tissue type;
the system wherein classifies patient biological samples as disease or healthy based on differential gene expression profiles of the fingerprint genes;
wherein the system stores the fingerprint gene information for each specific disease and cell/tissue type in a database that is referenced for diagnosing disease and classifying patients as disease or healthy;
wherein the system performs efficient dimensionality reduction of gene expression profile data of patients' biological samples to reduce computational burden, and enhance feature selection through identification of fingerprint genes to accurately diagnose and treat diseases;
wherein the system provides near real time (4-5 hours) diagnosis of disease at early stages for prescreening disease, and identification of feature genes that are specific to the disease, the stage of the disease, and the type of cell/tissue;
the system wherein confirms existence of disease if any, and identifies the stage of the disease in the patient;
the system wherein will enable a prompt therapy to be applied in response to the diagnosis.

### Claim 2 (depends on 1)

The method of claim 1 wherein the multilayer PCA system comprises:
a plurality of PCA layers that analyzes gene expression profiles of disease specific biological samples of patients and healthy subjects, for selection of optimal features called fingerprint genes;
wherein the plurality of layers analyzes gene expression profiles of biological samples of patients and healthy subjects, characteristic of different diseases, across several stages of disease and across different cell/tissue types;
the system wherein the fingerprint genes selected from the plurality of layers is tabulated in a database as a lookup table called fingerprint gene database;
wherein the plurality of layers is used for classification of patient biological sample as disease or healthy using the selected fingerprint genes;
wherein the plurality of layers is used for identifying stage of a disease;
wherein a therapy can be applied in response to the diagnosis.

### Claim 3 (depends on 2)

The method of claim 2 wherein the selection of the fingerprint genes through the plurality of PCA layers comprises of:
performing a linear transformation on the matrix of genes by disease specific patient biological samples to create principal component vectors from which a matrix of gene scores is created using the leading principal component vector, to select genes for Gene Pool A;
performing a linear transformation on the matrix of disease specific patient biological samples by genes, to create principal component vectors from which the leading principal component vector is used to select genes for Gene pool B;
wherein the genes common to pool A and pool B are selected as disease specific fingerprint genes to be tabulated in the fingerprint gene database for use in disease classification and early disease diagnosis;
wherein the early disease diagnosis will enable appropriate therapeutic treatment for the patient.

### Claim 4 (depends on 3)

The method of claim 3 wherein the selection of genes comprises:
genes in Gene Pool A selected based on their absolute gene scores greater than an adaptive threshold that can be variable between 60% to 85% of the maximum gene score;
genes in Gene Pool B selected based on absolute value of coefficients greater than 1/(sqrt(number of gene coefficients in the principal component));
genes common to Gene Pool A and Gene Pool B are selected as fingerprint genes.

### Claim 5 (depends on 2)

The method of claim 2 wherein the plurality of layers in the multilayer PCA further comprises of:
selection of fingerprint genes using PCA-Layer 1 and PCA-Layer 2;
classification of fingerprint genes as upregulated or down regulated using PCA-Layer 3;
classification and/or validation of patient biological sample as disease or healthy using PCA-Layer 4.

### Claim 6

A method of creating a fingerprint gene database comprising:
acquiring gene expression data of disease and healthy human subject biological samples;
determining optimal features called fingerprint genes specific to a disease, stage of the disease, and disease cell/tissue type by implementing the multilayer PCA classifier on the gene expression data;
tabulating the expression level of each fingerprint gene as upregulated or downregulated in the disease biological sample;
tabulating the fingerprint genes as a fingerprint gene library of disease-specific genes to be used for future disease classification;
tabulation of fingerprint gene data matrix of disease specific fingerprint genes and corresponding disease subject samples;
wherein the fingerprint gene library and the fingerprint gene data matrix together constitute the fingerprint gene database which can be used for disease diagnosis.

### Claim 7 (depends on 6)

The method of claim 6 wherein the fingerprint gene database used to perform patient disease pre-screening and early disease detection comprises:
execution of the multilayer PCA classifier using each set of disease specific fingerprint genes from fingerprint gene database on patient gene expression profile data;
wherein the PCA classifier is implemented on patient biological sample data using as many disease specific fingerprint genes that are tabulated in the fingerprint gene database to identify the disease if present in the patient;
wherein the early detection of disease can enable therapeutic treatment for the patient.

### Claim 8 (depends on 6)

The method of claim 6 wherein the fingerprint gene database used to confirm existing disease condition and identify the stage of disease in the patient further comprises:
execution of the multilayer PCA classifier using disease specific fingerprint genes from fingerprint gene database on patient gene expression profile data;
execution of the multilayer PCA classifier using disease stage specific fingerprint genes from fingerprint gene database on patient gene expression profile data.

### Claim 9

A method of classifying high dimensional data separable into two classes based on the statistical property of the classes of data using the multilayer PCA classifier by identifying fingerprint features in the two classes of data
