# US 2022/0230764 A1 · System for pathogenicity prediction of genomic variant using knowledge transfer

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

- Number: US 2022/0230764 A1
- Document ID: US20220230764A1
- Status: Published application
- Kind: A1
- Application number: 17/644,339
- Filed: 2021-12-15
- Published: 2022-07-21
- Earliest claimed: 2021-01-19
- Assignees: 3Billion
- Applicants: 3Billion
- Inventors: Kyoungyeul Lee; Dhong-gun Won
- Web page: https://patentgrep.com/patent/US20220230764A1

## CPC

- G16H 50/20: for computer-aided diagnosis, e.g. based on medical expert systems
- G16H 70/60: relating to pathologies
- G06N 3/0442: characterised by memory or gating, e.g. long short-term memory [LSTM] or gated recurrent units [GRU]
- G06N 3/045: Combinations of networks
- G06N 3/0464: Convolutional networks [CNN, ConvNet]
- G06N 3/08: Learning methods
- G06N 3/09: Supervised learning
- G06N 3/096: Transfer learning
- G06N 5/02: Knowledge representation; Symbolic representation
- G16B 15/20: Protein or domain folding
- G16B 20/50: Mutagenesis
- G16B 30/10: Sequence alignment; Homology search
- G16B 40/20: Supervised data analysis
- G16H 50/70: for mining of medical data, e.g. analysing previous cases of other patients
- Y02A 90/10: Information and communication technologies [ICT] supporting adaptation to climate change, e.g. for weather forecasting or climate simulation
- G16H: HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
- G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G16B: BIOINFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR GENETIC OR PROTEIN-RELATED DATA PROCESSING IN COMPUTATIONAL MOLECULAR BIOLOGY
- Y02A: TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE

## Abstract

The present disclosure provides a system for Pathogenicity prediction of a genomic variant using knowledge transfer, wherein the system is configured to learn an artificial neural network model using virtual genomic variant data generated from evolutionary conservation data, and learn actual genomic variant data by transferring knowledge by sharing weight values of a hidden layer extracted from the artificial neural network model to the artificial neural network model.

## Claims

### Claim 1

A system for Pathogenicity prediction of a genomic variant using knowledge transfer, wherein the system is configured to learn an artificial neural network model using virtual genomic variant data generated from evolutionary conservation data, and learn actual genomic variant data by transferring knowledge by sharing weight values of a hidden layer extracted from the artificial neural network model to the artificial neural network model.

### Claim 2 (depends on 1)

The system of claim 1, wherein the system comprises:
a virtual genomic variant data generation unit for generating the virtual genomic variant data from the evolutionary conservation data;
a virtual variant learning unit for learning an artificial neural network model using the virtual genomic variant data;
an actual variant learning unit for learning an artificial neural network model using the actual genomic variant data; and
a weight extraction unit for obtaining weight values of a hidden layer of the artificial neural network model when the virtual variant learning unit or the actual variant learning unit learns the artificial neural network model,
wherein the actual variant learning unit applies the extracted weight value to the hidden layer when learning an artificial neural network model.

### Claim 3 (depends on 1)

The system of claim 1, wherein the virtual genomic variant data generation unit comprises:
an evolutionary conservation data generation unit for generating the evolutionary conservation data including an evolutionary conservation feature by using multiple sequence alignment (MSA) from target protein sequence data and a plurality of similar pieces of protein sequence data; and
a virtual pathogenic variant determination unit for generating each of virtual pathogenic genomic variant data and virtual non-pathogenic genomic variant data according to preset criteria from the evolutionary conservation feature.

### Claim 4 (depends on 3)

The system of claim 3, wherein the evolutionary conservation feature is a frequency of amino acids found in a corresponding residue.

### Claim 5 (depends on 3)

The system of claim 3, wherein the multiple sequence alignment is performed by a BLAST algorithm or an HHBLits algorithm.

### Claim 6 (depends on 3)

The system of claim 3, wherein the evolutionary conservation data is an N×21-dimensional feature matrix, and the N is an arbitrary number corresponding to a length of an amino acid sequence.

### Claim 7 (depends on 2)

The system of claim 2, wherein the actual genomic variant data comprises actual pathogenic genomic variant data and actual non-pathogenic genomic variant data.

### Claim 8 (depends on 2)

The system of claim 2, wherein the knowledge transfer comprises transfer learning and multi-task learning.

### Claim 9 (depends on 8)

The system of claim 8, wherein in the transfer learning, after the virtual variant learning unit learns the virtual genomic variant data using an artificial neural network model, weight values extracted by the weight extraction unit is used by the actual variant learning unit.

### Claim 10 (depends on 8)

The system of claim 8, wherein in the multi-task learning, the respective weight values extracted from the virtual variant learning unit and the actual variant learning unit are alternately applied to a hidden layer of an artificial neural network model.

### Claim 11 (depends on 2)

The system of claim 2, wherein the hidden layer is an initial layer of the artificial neural network model.

### Claim 12 (depends on 2)

The system of claim 2, wherein further comprising a pathogenic determination unit that determines pathogenicity of a target genomic variant using an artificial neural network model learned by the actual variant learning unit.
