1. Deep Learning Analysis of Gene Expression Data for Breast Cancer Classification
AS Y.P. Manawadu

2. Problem Background Breast Cancer

3. Breast Cancer “ A kind of cancer that develops from breast tissue
which usually starts off in the inner lining of milk ducts
or the lobules. ”  NCI-
one of the major causes of death for women in the last decade
occurs in both men and women, although male breast cancer is rare.
About one in eight women are diagnosed with breast cancer during their lifetime
Men don't have breasts, so how can they get breast cancer? The truth is that boys and girls, men and women all have breast tissue. The various hormones in girls' and women's bodies stimulate the breast tissue to grow into full breasts. Boys' and men's bodies normally don't make much of the breast-stimulating hormones
U.S. Department of Health and Human Services
 National Institutes of Health
 National Cancer Institute
 USA.gov
There’s a good chance of recovery if it’s detected in its early stages.

4. Breast Cancer
Breast cancer is a highly heterogeneous disease due to its 
diverse morphological features
variable clinical outcome
 response to different therapeutic options
Therefore, it is necessary to devise a clinically meaningful classification of the disease, which has to be
scientifically sound
clinically useful
widely reproducible.
morphological features
Morphology is a branch of biology dealing with the study of the form and structure of organisms and their specific structural features.
Morphological features of Breast cancer :  tumor-size,  inv-nodes, node-caps. 4

5. Breast Cancer Classification
Classification based on pathology
classified according to its cellular structure and microanatomy.
Classification according to grade
grade 1 (well differentiated) – the cancer cells look most like normal cells and are usually slow-growing
grade 2 (moderately differentiated) – the cancer cells look less like normal cells are growing faster
grade 3 (poorly differentiated) – the cancer cells look most changed and are usually fast-growing.
Each of these schemes classify the cancers based on different criteria and serve a different purpose. 

6. Breast Cancer Classification
3. Classification based on stage of cancer
TNM staging that takes into account the Tumor size, lymph Node involvement and Metastasis or spread of the cancer.
4. Classification based on Protein & gene status
takes into account the estrogen receptor (ER), progesterone receptor (PR) and HER2 proteins.
Once the status of these proteins is known, prognosis can be predicted and certain novel therapies may be chosen for treatment.
Each of these schemes classify the cancers based on different criteria and serve a different purpose. 
All breast cancers these days are tested for expression, or detectable effect, of the estrogen receptor (ER), progesterone receptor (PR) and HER2 proteins.

7. Breast Cancer Types
Endocrine/Hormone receptor-positive (estrogen or progesterone receptors)
About 80% of all breast cancers are “ER-positive.”   
cancer cells grow in response to the hormone estrogen
About 65% of these are “PR-positive.”
grow in response to hormone, progesterone.
HER2-positive
cells make too much of a protein known as HER2. 

8. Breast Cancer Types Triple negative :
they don’t have estrogen and progesterone receptors and don’t overexpress the HER2 protein
Most breast cancers associated with the gene BRCA1 are triple negative.
Triple positive : positive for estrogen receptors, progesterone receptors, and HER2
Possible treatments are different.
Based on hormons and genes
no targeted therapies have been developed to help prevent cancer returning in women with triple-negative breast cancer.

9. Problem Background Gene Expression Data

10. Gene Expression Data rows represent genes,
=The raw microarray data are images, which have to be transformed into gene expression matrices tables where
rows represent genes,
columns represent various samples such as tissues or experimental conditions
numbers in each cell characterize the expression level of the particular gene in the particular sample.
Microarray data allow monitoring of gene expression for tens of thousands of genes in parallel and are already producing huge amounts of valuable data.
Numbers in each cell characterize the expression level of the particular gene in the particular sample.

11. Why Gene Expression Data for breast cancer classification ?
Gene expression patterns were found to be strongly associated with estrogen receptor (ER) status and moderately associated with grade.
(Marc J. van de Vijver et al, 2002)
Other data sets : Morphological features
Comprehensive gene expression patterns generated from cDNA microarrays were correlated with detailed clinico-pathological characteristics and clinical outcome in group of breast cancer patients. 
(Christos Sotiriou,2003)

12. Problem Background Deep Learning

13. Deep Learning
A new area of Machine Learning research, which has been introduced with the objective of
moving Machine Learning closer to one of its original goals: Artificial Intelligence.

14. Why Deep Learning ?
perform better for classification than other traditional Machine Learning methods, because:
deep learning methods include multi layer processing with less time and better accuracy performance.
Sub sampling layers give better result ,by use of CNN and auto-encoders.
With the increase number of auto encoders, the accuracy increases. Similarly increase number of sub sampling too gives the better.

15. Problem , Objectives & Methodology

16. Problem Definition
Numerous researches available for the analysis of different data matrices using different algorithms related to breast cancers.
No any research done using deep learning applied to gene expression data for classifying breast cancer.

17. Objectives
Find whether deep Learning applied to gene expression data can successfully classify breast cancer than the other algorithms.
Comparative analysis of different deep Learning algorithms applied to gene expression data for classifying breast cancer.
used in machine learning for the classification of Breast cancer.

18. Methodology Select suitable deep learning algorithms.
Select appropriate datasets.
Carryout the analysis by employing the selected algorithms.
Generate statistics for each of the algorithm.
Device conclusions on the performance of each algorithm, comparatively.
Select suitable deep learning algorithms, which can be applied to classifying gene expression data.
Select appropriate datasets and construct a database to be used for the analysis.
Carryout the analysis employing each of the algorithms.
Generate statistics of each of the outputs.
Device conclusions on the performance of each algorithm, comparatively.

19. TimeLine Study about Deep Learning and its algorithms
Researches related to breast cancer, Gene Expression data & Deep Learning
Datasets used in researches related to deep learning & compare those datasets with gene expression data

20. Related Work & References

21. Related Work
Breast cancer classification and prognosis based on gene expression profiles from a population-based study
(Christos Sotiriou et al. , 2003)
 Hierarchical cluster analysis
Gene expression profiling in breast cancer: classification, prognostication, and prediction (Prof. Jorge S Reis-Filho et al. , 2011)
a molecular classification system and prognostic multigene classifiers based on microarrays or derivative technologies.
Gene Expression Profiling in Breast Cancer: Understanding the Molecular Basis of Histologic Grade To Improve Prognosis 
(Christos Sotiriou et al. ,2006)
Kaplan–Meier analysis
1,3,4,5- request for dataset sent
6- error in sending the request
2- could not send
gene expression profiling is the measurement of the activity (the expression) of thousands of genes at once, to create a global picture of cellular function.
Prognosis= the prospect of recovery as anticipated from the usual course of disease or peculiarities of the case.

22. Related Work
Distinct molecular mechanisms underlying clinically relevant subtypes of breast cancer: gene expression analyses across three different platforms. (Therese Sorlie et al. , 2006)
hierarchical clustering and centroid correlation analysis
Classification of human breast cancer using gene expression profiling as a component of the survival predictor algorithm. (Gennadi V. Glinsky et al. , 2004)
microarray expression profiling and quantitative reverse transcription using Kaplan-Meier analysis.
Gene expression patterns of breast carcinomas distinguish tumor subclasses with clinical implications. (Therese Sorlie et al. , 2001)
Microarray Analysis by Hierarchical Clustering.

23. Related Work
Breast cancer classification using deep belief networks. (Ahmed M. et al. , 2016)
Morphological features dataset.
99.68%

24. References
Christos Sotiriou, Soek-Ying Neo and Lisa M. ‘Breast cancer classification and prognosis based on gene expression profiles from a population-based study’ , The National Academy of Sciences , vol. 100 no. 18,
Prof. Jorge S Reis-Filho and FRCPath. ‘Gene expression profiling in breast cancer: classification, prognostication, and prediction’ , The Lancet , vol no. 9805,
Christos Sotiriou, Pratyaksha Wirapati and Sherene Loi. ‘Gene Expression Profiling in Breast Cancer: Understanding the Molecular Basis of Histologic Grade To Improve Prognosis’ , Journal of the National Cancer Institute, vol. 98 no. 4,
Marc J. van de Vijver et al. , A Gene-Expression Signature as a Predictor of Survival in Breast Cancer, N Engl J Med, Vol. 347 No. 25, 2002

25. References
Therese Sorlie, Yulei Wang, Chunlin Xiao and Hilde Johnsen. ‘Distinct molecular mechanisms underlying clinically relevant subtypes of breast cancer: gene expression analyses across three different platforms’ , BMC Genomics, vol. 7 no. 127,
Gennadi V. Glinsky, Takuya Higashiyama and Anna B. Glinskii . ’ Classification of human breast cancer using gene expression profiling as a component of the survival predictor algorithm’, Clinical Cancer Research, vol. 10 no. 7,
Therese Sorlie and Charles M. Peroua. ‘Gene expression patterns of breast carcinomas distinguish tumor subclasses with clinical implications’ , The National Academy of Sciences, vol. 98 no. 19, 2001.
Ahmed M. Abdel-Zaher, Ayman M. Eldeib. ‘Breast cancer classification using deep belief networks’. Expert Systems With Applications , vol.46 no. 139, 2016.

26. Thank You Any Questions ?