IDENTIFYING CANCER GENE USING

NEURAL NETWORKS

Presenter Details
717821L215-HEMANANTH V
717821L221-KAVIN S
717821L224-KISHORE S
717821L230-MITHRA N R
INTRODUCTION
• Cancer gene identification is a vital aspect of cancer research focused on identifying specific genes associated
with the initiation, progression, or susceptibility to various types of cancer. Understanding the genetic factors
underlying cancer allows researchers and clinicians to develop targeted therapies, enhance early detection
methods, and personalize treatment approaches. In recent years, advanced computational methods,
particularly machine learning and deep learning algorithms, have played a significant role in cancer gene
identification. These algorithms analyze large-scale genomic datasets to identify patterns, mutations, and
associations that may not be immediately apparent through traditional methods. Convolutional Neural
Networks (CNNs), Recurrent Neural Networks (RNNs), and other deep learning architectures are employed
to uncover intricate relationships within genomic data. The primary aim of this research is to develop a user-
friendly interface for identifying cancer genes through a deep learning model. The implementation utilizes
CNN, a machine learning model specifically designed for computer vision and image processing tasks, with a
dedicated focus on cancer gene detection. A CNN model is constructed using the training data, and finally, the
test data is classified using CNN models known for their capacity to yield highly accurate classifications. This
system aims to streamline the process of cancer gene identification, offering an accessible interface driven by
advanced deep learning techniques.

7/30/2024 2
NEED FOR CURRENT WORK

• Cancer genetics is inherently complex, involving numerous genetic and molecular factors. Traditional
methods often struggle to capture the intricate patterns and relationships in large-scale genomic data.
Neural networks, particularly Convolutional Neural Networks (CNNs), offer advanced capabilities to
analyze these complex data sets effectively.
• Traditional cancer gene identification methods may be limited in their ability to handle high-
dimensional data and complex relationships between genes. Advanced machine learning models like
CNNs can provide better accuracy and uncover hidden patterns that traditional approaches might miss.
• Early detection of cancer is crucial for effective treatment and better outcomes. The proposed neural
network model aims to provide highly accurate predictions of cancer-related genes, which can facilitate
earlier diagnosis and more personalized treatment plans.

7/30/2024 3
LITERATURE SURVEY
• Identifying cancer genes is a crucial aspect of cancer research, and various techniques and approaches
have been developed to uncover genes that play a role in cancer initiation, progression, and
development. Here are some existing techniques for cancer gene identification. various machine
learning and deep learning techniques have used in cancer gene identification. The detection accuracy is
still in consideration.

• Detecting cancer genes involves a comprehensive analysis of genetic data to identify specific genes
associated with the initiation, development, or susceptibility to cancer.

• In their work, Won et al. [4] introduced a method employing an ensemble of neural network classifiers
derived from negatively correlated characteristics for precise cancer categorization. The effectiveness of
the proposed technique was evaluated using three benchmark datasets. Experimental results
demonstrated that the ensemble classifier utilizing negatively correlated characteristics achieved the
highest recognition rate among these benchmark datasets.

7/30/2024 4
PROBLEM STATEMENT

• Identifying cancer-related genes using neural networks is a complex and challenging task that involves
analyzing large-scale genomic and transcriptomic data. Neural networks, a subset of machine learning
models, can be employed to discover patterns and relationships within biological data that may be
indicative of cancer-related genetic variations. Now a day’s cancer gene identification is a complicated
and important step in medical field. Neural networks are helpful in cancer gene identification due to
their ability to analyze complex patterns and relationships within large-scale genomic and clinical data.
The convoluted nature of cancer involves numerous genetic and molecular factors that can be
challenging for traditional methods to fully comprehend. Neural networks, particularly deep learning
models, excel at learning hierarchical representations and capturing intricate patterns in data. The
proposed work utilizes convolution neural network to identify cancer genes. The proposed work utilizes
python tool to detect the cancer gene. The main objective of the proposed system is to provide a user-
friendly interface for Identifying Cancer Gene detection using deep learning model. The proposed
application first captures RNA-Seq data from The Cancer Genome Atlas (TCGA) .Then data is
extracted using the data extraction process and Build CNN model using train data. Finally test data is
classify using CNN models are known to produce more accurate classifications.

7/30/2024 5
EXISTING METHOD

• While genomic sequencing provides detailed information about genetic variations, it often requires
extensive analysis and interpretation. Traditional methods may not efficiently handle the vast amounts of
data generated or identify subtle patterns related to cancer.
• Approaches like functional genomics focus on understanding gene functions and interactions but may
not always provide comprehensive insights into cancer gene identification due to the complexity and
variability of cancer biology.
• Many advanced machine learning models, including CNNs, are often criticized for their lack of
interpretability. Understanding how these models arrive at specific predictions can be challenging,
which is a critical consideration for clinical applications.
• The effectiveness of existing methods is heavily dependent on the quality and diversity of the data used
for training. Incomplete or biased datasets can lead to inaccurate predictions and limit the
generalizability of the models.

7/30/2024 6
TIMELINE
1994
• Hu et al.: Introduced a cancer classification technique combining supervised and
unsupervised learning methods.
2003
• Won et al.: Developed an ensemble of neural network classifiers derived from negatively
correlated features for cancer categorization.
2005
• Xu et al.: Proposed a hybrid method using a probabilistic neural network (PNN) and Particle
Swarm Optimization (PSO) for gene selection and cancer survival prediction.
2006
• Bevilacqua et al.: Presented an approach using an artificial neural network (ANN) optimized
through a multi-objective genetic algorithm for cancer classification.

7/30/2024 7
CONCLUSION

• The proposed neural network model for cancer gene detection has been implemented using python tool
and its shows the best results with providing highest accuracy in classification; it gives the 96.04%
result as compare to other classifiers. The proposed Neural network model helps to people to be checked
for cancer disease and thus detecting the disease at an early stage. This robust neural network model
serves as a valuable tool for individuals undergoing cancer screening, facilitating early detection of the
disease. The high accuracy implies the effectiveness of neural networks in aiding oncologists in cancer
diagnosis. Such predictions not only enable doctors to plan more effective treatments but also empower
patients with early awareness of their health status. Furthermore, extracting additional image features
could enhance the classification process, and exploring various combinations of existing features holds
potential for more accurate medical data classification. Overall, the utilization of neural networks in
cancer gene detection holds promise for advancing our understanding of the genetic basis of cancer,
fostering personalized medicine, and ultimately improving diagnostic accuracy in the pursuit of effective
cancer treatments.

7/30/2024 8
REFERENCE

1. Hong-Hee Won, Sung-Bae Cho. Paired neural network with negatively correlated features for cancer
classification in DNA gene expression profiles; In Proceedings of the International Joint Conference on
Neural Networks; 3; 2003; p. 1708 – 1713.
2. Hu Y., Ashenayi K., Veltri, R., O'Dowd G., Miller G., Hurst R. Bonner. A comparison of neural
network and fuzzy c-means methods in bladder cancer cell classification; In Proceedings of IEEE
World Congress on Computational Intelligence; 6; 1994; p. 3461 – 3466.
3. Bevilacqua, V., Mastronardi, G., Menolascina, F., Pannarale P., Pedone. A Novel Multi-Objective
Genetic Algorithm Approach to Artificial Neural Network Topology Optimisation; The Breast Cancer
Classification Problem; In Proceedings of International Joint Conference on Neural Networks (IJCNN
'06);2006; p. 1958 – 1965.

7/30/2024 9
 THANK YOU

7/30/2024 10