# US 8,788,444 B2 · Data analysis method and system

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

- Number: US 8,788,444 B2
- Document ID: US8788444B2
- Status: Granted
- Kind: B2
- Application number: 13/125,054
- Filed: 2009-10-20
- Published: 2011-11-17
- Granted: 2014-07-22
- Earliest claimed: 2008-10-20
- Assignees: Nottingham Trent University
- Applicants: Graham Ball; Lee Lancashire
- Inventors: Graham Ball; Lee Lancashire
- Web page: https://patentgrep.com/patent/US8788444B2

## CPC

- G06N 3/105: {Shells for specifying net layout}
- C12Q 1/6886: for cancer (immunoassay for cancer G01N33/575)
- G16B 25/10: Gene or protein expression profiling; Expression-ratio estimation or normalisation
- G16B 40/00: ICT specially adapted for biostatistics; ICT specially adapted for bioinformatics-related machine learning or data mining, e.g. knowledge discovery or pattern finding
- G16B 40/20: Supervised data analysis
- G16B 25/00: ICT specially adapted for hybridisation; ICT specially adapted for gene or protein expression
- Y02A 90/10: Information and communication technologies [ICT] supporting adaptation to climate change, e.g. for weather forecasting or climate simulation
- G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- C12Q: MEASURING OR TESTING PROCESSES INVOLVING ENZYMES, NUCLEIC ACIDS OR MICROORGANISMS (immunoassay G01N33/53); COMPOSITIONS OR TEST PAPERS THEREFOR; PROCESSES OF PREPARING SUCH COMPOSITIONS; CONDITION-RESPONSIVE CONTROL IN MICROBIOLOGICAL OR ENZYMOLOGICAL PROCESSES
- 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 invention relates to the analysis of data to identify relationships between the input data and one or more conditions. One method of analyzing such data is by the use of neural networks which are non-linear statistical data modelling tools, the structure of which may be changed based on information that is passed through the network during a training phase. A known problem that affects neural networks is the issue of overtraining which arises in overcomplex or overspecified systems when the capacity of the network significantly exceeds the needed parameters. The present invention provides a method of analyzing data using a neural network with a constrained architecture that mitigates the problems associated with the prior art.

## Claims

### Claim 1

A method of determining a relationship between input data and one or more conditions comprising the steps of:
receiving input data categorised into one or more predetermined classes of condition;
training an artificial neural network with the input data, the artificial neural network comprising an input layer having one or more input nodes arranged to receive input data; a hidden layer comprising two or more hidden nodes, the nodes of the hidden layer being connected to the one or more nodes of the input layer by connections of adjustable weight; and, an output layer having an output node arranged to output data related to the one or more conditions, the output node being connected to the nodes of the hidden layer by connections of adjustable weight;
determining relationships between the input data and the one or more conditions
wherein the artificial neural network has a constrained architecture in which
(i) the number of hidden nodes within the hidden layer is constrained; and,
(ii) the initial weights of the connections between nodes are restricted and
wherein the input data comprises data pairs, each data pair being categorised into the one or more conditions and comprising a parameter and associated parameter value, the training and determining steps comprising:
(i) selecting a parameter within the input data, training the artificial neural network with corresponding parameter values and recording artificial neural network performance;
(ii) repeating for each parameter within the input data;
(iii) determining the best performing parameter in the input data;
(iv) repeating steps (i) to (iii), each repetition adding one of the remaining parameters to the best performing combination of parameters, until artificial neural network performance is not improved.

### Claim 2 (depends on 1)

A method of determining a relationship between input data and one or more conditions as claimed in claim 1, wherein the number of nodes in the hidden layer is in the range two to five.

### Claim 3 (depends on 2)

A method of determining a relationship between input data and one or more conditions as claimed in claim 2, wherein there are two hidden nodes in the hidden layer.

### Claim 4 (depends on 1)

A method of determining a relationship between input data and one or more conditions as claimed in claim 1, wherein the initial weights of connections between nodes have a standard deviation in the range 0.01 to 0.5.

### Claim 5 (depends on 4)

A method of determining a relationship between input data and one or more conditions as claimed in claim 4, wherein the initial weights of connections between nodes have a standard deviation of 0.1.

### Claim 6 (depends on 1)

A method of determining a relationship between input data and one or more conditions as claimed in claim 1, wherein the input data is grouped into a plurality of samples, each sample having an identical selection of data pairs.

### Claim 7 (depends on 6)

A method of determining a relationship between input data and one or more conditions as claimed in claim 6, wherein the training step comprises:
(i) selecting in a first selecting step the same parameter in each sample;
(ii) training the artificial neural network with the parameter values associated with the selected parameter;
(iii) recording the artificial neural network performance for the selected parameter;
(iv) repeating the selecting and recording steps for each parameter in turn.

### Claim 8 (depends on 7)

A method of determining a relationship between input data and one or more conditions as claimed claim 7, wherein the determining step further comprises:
(i) ranking the performance of the artificial neural network for each selected parameter based on their recorded performance, and;
(ii) selecting, in a second selecting step, the best performing parameter.

### Claim 9 (depends on 8)

A method of determining a relationship between input data and one or more conditions as claimed in claim 8, wherein the training step further comprises:
(i) selecting, in a further selecting step, a parameter from the remaining parameters in conjunction with the best performing parameter or parameters from the previous selecting step;
(ii) training the artificial neural network with the parameter values associated with the selected parameters;
(iii) recording, in a further recording step, the artificial neural network performance for the selected parameters, and;
(iv) repeating the further selecting and recording steps for each of the remaining parameters in turn.

### Claim 10 (depends on 9, 10)

A method of determining a relationship between input data and one or more conditions as claimed in claim 9, wherein the training step further comprises repeating steps (i)-(iv) of claim 10 until no further substantial performance increase is gained.

### Claim 11 (depends on 9)

A method of determining a relationship between input data and one or more conditions as claimed in claim 9, wherein, prior to the repeating step, the determining step comprises ranking the performance of the artificial neural network for each selected parameter arrangement based on their recorded performance and selecting the best performing parameters.

### Claim 12 (depends on 9)

A method of determining a relationship between input data and one or more conditions as claimed in claim 9, wherein each time the number of input parameters is increased the number of input nodes is increased by one node.

### Claim 13 (depends on 1)

A method of determining a relationship between input data and one or more conditions as claimed in claim 1, wherein the parameters represent genes and the parameter values represent gene expression data.

### Claim 14 (depends on 1)

A method of determining a relationship between input data and one or more conditions as claimed in claim 1, wherein the parameters represent proteins and the parameter values represent activity function.

### Claim 15 (depends on 1)

A method of determining a relationship between input data and one or more conditions as claimed in claim 1, wherein the artificial neural network is trained using random sample cross validation.

### Claim 16 (depends on 15)

A method of determining a relationship between input data and one or more conditions as claimed in claim 15, wherein input data is randomly divided into three subsets, a first subset for training the network, a second subset to assess the network during the training process and a third subset to validate the trained network on unseen data.