基因芯片对癌症诊断及分类应用.ppt

Class Discovery: defining previously unrecognized tumor subtypes Class Prediction: assignment of particular tumor samples to already-defined classes Given bone marrow samples: Which cancer classes are present among sample? How many cancer classes? 2, 4? Given samples are from leukemia patients, what type of leukemia is each sample (AML vs ALL)? supervised分析： 1. 以已知患有白血病的患者为training set 2.用病人的切片进行microarry反应，建立了基因图谱 3.根据图谱选出在两类中具有表达差异的基因作为predictor set 4.试验样本testing set在predictor set上micoarray反应 5.再根据testing set中不同的predictor基因表达情况将其分类 unsupervised分析: 1．如果某中疾病的分类未知，可将training set的切片在microarry上进行全基因表达 2．根据表达相似和差异性进行认为分类，本图中是分为两类 3．然后同supervised分析，将testing set 归到人为分类的各类中 Cancer of bone marrow骨髓癌 Myelogenous or lymphocytic, acute or chronic Acute Myelogenous Leukemia (AML) vs Acute Lymphocytic Leukemia (ALL) Marrow cannot produce appropriate amount of red and white blood cells Anemia - weakness, minor infections; blood cell deficiency - easy bruising AML: 10,000 new adult cases per year ALL: 3,500/2,400 new adult/child cases per year 4.research in Hematologic malignancies：Leukemia非白血性白血病 Leukemia: Treatment expected outcome Diagnosis via highly specialized laboratory ALL: 58% survival rate AML: 14% survival rate Treatment: chemotherapy, bone marrow transplant ALL: corticosteroids, vincristine, methotrexate, L-asparaginase甲氨喋呤 AML: daunorubicin, cytarabine阿糖胞苷 Correct diagnosis very important for treatment options and expected outcome!!! Microarray could provide systematic diagnosis option 38 bone marrow samples (27 AML, 11 ALL) 6817 human gene probes Leukemia: Data set General method for prediction issues: Determine if there are any genes with expression pattern correlated with class distinction Given a collection of known samples, create a class predictor Test validity of predictors Prediction: Methodology C* idealized random pattern Find genes Make predictor Test predictors Prediction: Determine significant genes Prediction: Terminology for class predictor Fi