Classification of Human Acute Myeloid Leukemia & Acute Lymphoid Leukemia by Gene Expression Monitoring Affiliation: Edward Waters College By: Myschelle Jones

Table of Contents
Page 2-Table of Contents
Page 3-Abstract
Page 4-5-Introduction
Page 6-Process of Distinguishing AML and ALL
Page 7-Results
Page 8-Conclusion
Page 9-Discussion
Pages 10-12-References

Abstract
Cancer classification using gene expression data is known to contain the keys for addressing the fundamental problems relating to cancer diagnosis and drug discovery. Cancer research is one of the major research areas in the medical field. Accurate prediction of different tumor types has great value in providing better treatment and toxicity minimization on the patients. Performing gene selection helps to reduce data size. More importantly, gene selection removes a large number of irrelevant genes, which improves the classification accuracy.

Introduction
Leukemia is a type of blood cancer that is “often fatal”.[1] The white blood cells displace normal blood, leading to infection, shortage of red blood cells anemia, bleeding, and other disorders. Certain types of childhood leukemia respond well to treatment, “which includes drugs, chemotherapy, and radiotherapy”. [2] There are two types of leukemia that will be discussed, acute lymphocytic leukemia and acute myelogenous leukemia. Acute lymphocytic leukemia is the most common “type of cancer in children”. [1] Acute lymphocytic leukemia (ALL) is a type of cancer of the “blood and bone marrow — the spongy tissue inside bones where blood cells are made”. [1] Acute myelogenous leukemia denotes “diseases rapid progression”. [3] The word "acute" in acute lymphocytic leukemia comes from the fact that the disease progresses rapidly and affects immature blood cells, rather than mature ones. The "lymphocytic" in acute lymphocytic leukemia refers to the white blood cells called lymphocytes, which ALL affects. Acute lymphocytic leukemia is also known as acute lymphoblastic leukemia and acute childhood leukemia.
ALL is the most common type of cancer in children, and “treatments result in a good chance for a cure”. [4] ALL can also occur in adults, though the “prognosis is not as optimistic”. [3] Distinguishing ALL from AML is critical for successful treatment; “chemotherapy regimens for ALL generally contain corticosteroids, vincristine, methotrexate, and L-asparaginase, whereas most AML regimens rely on a backbone of daunorubicin and cytarabine”. [5] Although remission can be achieved using ALL therapy for AML (and vice versa), cure rates are “markedly diminished, and unwarranted toxicities are encountered”. [5]
Acute myelogenous leukemia is caused by “damage to the DNA of developing cells in the bone marrow”. [6]When this happens, blood cell production goes awry. The bone marrow produces” immature cells that develop into leukemic white blood cells called myeloblasts”. [7] These abnormal cells are “unable to function properly”, and they can build up and crowd out healthy cells. [8] Its not clear what causes the “DNA mutations that lead to leukemia”. [9] Radiation, exposure to “certain chemicals and some chemotherapy drugs may play a role”. [10]
The purpose of this project is to use the data mining method for distinguishing AML from ALL, using gene expression data. The use of gene expression data differs from the traditional approach of relying on morphological appearance. Distinguishing ALL from AML is “critical for successful treatment”. [11] Today, the challenge regarding cancer treatment has been to target specific therapies to distinct types of the disease, to maximize efficacy and minimize toxicity.

Materials and Methods Data was received in Microsoft Excel to use computer programs to distinguish AML from ALL. The following software was used to aid in the completion of the project: Microsoft Excel, Python, Weka, and Crimson Editor. The data was entered into Microsoft Excel and randomized using a random number generator and a sort option. Using Crimson Editor a python program was written to read in the data from Microsoft Excel in Weka format. Once the data was in Weka format it was opened in Weka and the class was chosen from which the data would be tested on. In Weka there are “numerous ways to test data”. [12] For the” learning and testing process” the j48 decision tree was chosen. [13] Along with the j48 tree 5 fold cross validation was used. Using cross validation the Weka software further “randomized the data and produced instant results that correlated with the data”. [14] The j48 tree was displayed to” show which of the genes in the test data proved” [15] important when distinguishing between ALL and AML.

Results
After running WEKA utilizing J48, the end results were substantially great. Moreover, the program was able to accurately execute the task of differentiating AML, Acute Myeloid Leukemia, and ALL, Acute Lymphoid Leukemia. The quantity of correctly classified data was approximately 90.2778 percent. This percentage allowed 9.7222 percent for incorrectly classifying AML, and ALL. Moreover, there were only seven mistakes out of the seventy-two genes provided. The program used a variety of factors to yield our final results: kappa statistics, mean absolute error, root mean squared error, relative absolute error, and root relative squared error. Furthermore, these are the circumstances in determining whether the gene is better used to depict the cancer as AML, or ALL, and why. Our classifier, J48, was effective” utilizing the data provided to yield accurate results”. [13] The F-measure (combination of precision and recall) for AML was eighty-seven percent, and ALL was ninety-two percent. There were two genes whose expression levels were significant in distinguishing AML and ALL. They are X95735 and U53468; these genes had the greatest differentiation according to WEKA. Gene, X95735 expression levels were 423.1915 for ALL and 3285.92 for AML. For gene, U53468, the expression levels were 32.48 for ALL and 89.7234 for AML. Pertaining to X95735, if the expression numbers were less than, or equal to 938; it was classified as ALL, and greater than, AML. For U53468, if the expression numbers were less than 143 it was AML, and greater than, ALL. Lastly, the confusion matrix pinpointed where our errors occurred, giving an exact account. Forty-two ALL cases were correctly diagnosed, and twenty-three AML were as well. However, when the patient had ALL it was said to be AML twice and lastly when the patient had AML, they were told ALL five times.

Conclusion
In conclusion, WEKA was able to complete and yield results. It was found that focusing on two genes, U53468 and X95735, out of seventy-two, AML and ALL was classified at a substantially good level. Moreover, there was approximately a ninety percent success rate. Imperfections were shown with the program by there being an inaccuracy rate of 9.722%. Moreover, there were seven instances out of seventy-two, where AML was not properly distinguished from ALL. The gene levels expressed in AML were very high compared to the gene levels expressed in ALL which were very low. It is better to focus on two genes than focusing on all seventy-two. This could be due to the variables used for distinguishing the two types of cancer. The only variable utilized was gene expression in lieu of the plethora of other factors that exist.

Discussion
Cancer classification using gene expression data is known to “contain the keys for addressing the fundamental problems relating to cancer diagnosis and drug discovery”. [18] Cancer research is one of the major research areas in the medical field. Accurate prediction of different tumor types has great value in providing better “treatment and toxicity minimization” on the patients. [19] Performing gene selection helps to reduce data size. More importantly, gene selection “removes a large number of irrelevant genes”, which improves the classification accuracy. [20]

X95735, known as Zyxin, an adhesion plaque protein and a part of a signal transduction pathway that mediates adhesion stimulated changes in gene expression, “plays a significant role in leukemia pathogenesis in AML but not in ALL”. [16] Zyxin also plays a critical role in mitosis. Its location in Homo sapiens is on chromosome 7 (7q32). U53468, known as ubiquinone oxidoreductase subunit B13, is also in Homo sapiens on chromosome 7. This gene transfers electrons from NADH to ubiquinone. Ubiquinone is an oil-soluble substance that is mostly present in the mitochondria. It is part of the electron transport chain and helps produce energy into the ATP form. Ninety-five percent of the energy in the human body is produced this way so “organs that need the most energy (heart, liver, and kidney) have the most concentrations of ubiquinone”. [17]

References
1. "Adult Acute Myeloid Leukemia Treatment (PDQ®) - National Cancer Institute." Comprehensive Cancer Information - National Cancer Institute. Web. 12 July 2011. http://www.cancer.gov/cancertopics/pdq/treatment/adultaml/patient/page1>. 2. Golub, T.R. "Molecular Classification of Cancer: Class Discovery and Class Prediction by Gene Expression Monitoring.”Pages 531-537.
3. "Types of Leukemia." Mayo Clinic. Web. 13 July 2011. <http://www.mayoclinic.com>. Web. 13 July 2011. <http://www.ncbi.nlm.nih.gov>.
4. Dugdale III, David C., Jason R. Mason, and Torres Pines. "Acute Myeloid Leukemia: MedlinePlus Medical Encyclopedia." National Library of Medicine - National Institutes of Health. National Institutes of Health, 6 Sept. 2010. Web. 27 Jan. 2012. <http://www.nlm.nih.gov/medlineplus/ency/article/000542.htm>.
5. Jemal, A., A. Thomas, and Martin Abeloff. "Acute Myeloid Leukemia." Wikipedia, the Free Encyclopedia. Wikiepedia, 16 Jan. 2012. Web. 27 Jan. 2012. <http://en.wikipedia.org/wiki/Acute_myelogenous_leukemia>.
6. LeMaistre, C. F., Paul Shaughnessy, and Anthony S. Stein. "Acute Myelogenous Leukemia (AML)." Be The Match Marrow Registry. Marrow.org. Web. 31 Jan. 2012. <http://marrow.org/Patient/Disease_and_Treatment/About_Your_Disease/AML/Acute_Myelogenous_Leukemia_(AML).aspx>.
7. "Atlas Of Acute Leukemia (ALL, AML)." Cancer Information | Cancer Meds | Leukemia | Colon Cancer | Lung Cancer | HIV/AIDS | Hepatitis | Mesothelioma. Meds.com. Web. 31 Jan. 2012. <http://meds.com/leukemia/atlas/acute_leukemia.html>.
8. "Difference Between AML and ALL | Difference Between | AML vs ALL." Difference Between Similar Terms and Objects. Web. 31 Jan. 2012. <http://www.differencebetween.net/science/health/difference-between-aml-and-all/>.
9. "Acute Lymphoblastic Leukemia | The Leukemia & Lymphoma Society." The Leukemia & Lymphoma Society® - Official Website. Leukemia and Lymphoma Society. Web. 31 Jan. 2012. <http://www.lls.org/diseaseinformation/leukemia/acutelymphoblasticleukemia/>.
10. Appelbaum, F. M. "Acute Myeloid Leukemia - PubMed Health." Pubmed Health. 6 Sept. 2010. Web. 31 Jan. 2012. <http://www.ncbi.nlm.nih.gov/pubmedhealth/PMH0001569/>.
11. "AML Prognosis." Leukemia Home Page. Web. 31 Jan. 2012. <http://leukemia.emedtv.com/aml/aml-prognosis.html>.
12. "Acute Myelogenous Leukemia." 13 June 2011. Web. 31 Jan. 2012. <http://emedicine.medscape.com/article/197802-overview>.
13. "Markers For The Diagnosis Of Aml, B-all And T-all." Freshpatents.com: Patent Applications Updated Each Week, RSS, Keyword Monitoring. Web. 31 Jan. 2012. <http://www.freshpatents.com/Markers-for-the-diagnosis-of-aml-b-all-and-t-all-dt20080320ptan20080070793.php>.
14. "Adult Acute Myeloid Leukemia Treatment (PDQ®) - National Cancer Institute." Comprehensive Cancer Information - National Cancer Institute. Web. 31 Jan. 2012. <http://www.cancer.gov/cancertopics/pdq/treatment/adultAML/Patient/page1>.
15. "Difference between AML and CML." Web. 31 Jan. 2012. <http://www.differencebetween.net/science/health/difference-between-aml-and-cml-2/>.
16. Leukemia-Acute Myeloid in Children. Cancer.net. Web. 31 Jan. 2012. <http://www.cancer.net/patient/Cancer+Types/Leukemia+-+Acute+Myeloid+-+AML+-+Childhood?sectionTitle=Treatment>.
17. "Weka 3 - Data Mining with Open Source Machine Learning Software in Java." Computer Science Department, University of Waikato. Web. 31 Jan. 2012. <http://www.cs.waikato.ac.nz/ml/weka/>.
18. Weka Software Informer: Latest Version Download, News and Info about This Machine Learning Group, University of Waikato, Hamilton, NZ Program. It Allows You Can Extract Useful Information from Large Databases. Web. 31 Jan. 2012. <http://weka.software.informer.com/>.
19. "C4.5 Algorithm." Wikipedia, the Free Encyclopedia. Web. 31 Jan. 2012. <http://en.wikipedia.org/wiki/C4.5_algorithm>.
20. "Classification via Decision Trees in WEKA." DEPAUL UNIVERSITY, 2005. Web. 31 Jan. 2012. <http://maya.cs.depaul.edu/~classes/ect584/WEKA/classify.html>.