University of Washington Institute of Technology Tacoma, WA, USA Ecole des Hautes Etudes en Santé Publique Département Infobiostat Rennes, France Isabelle Bichindaritz
Purpose of this Talk Once upon a time … There was biology (~1800), and There were computers (~1920) Of their common interests was born bioinformatics (~1979) … Question: How can CBR contribute to bioinformatics research ? An example to microarray data analysis ICCBR '10
NCBI, 2004
Bioinformatics Challenges Frequent tasks in bioinformatics Similarity search in genetic sequences Microarray data analysis Macromolecule shape prediction Evolutionary tree construction Gene regulatory network mining ICCBR '10
Bioinformatics Challenges Microarray data analysis Microarrays are made from a collection of purified DNA’s. A drop of each type of DNA in solution is placed onto a specially- prepared glass microscope slide by an arraying machine. Please note that … … the human genome contains about 30,000 genes. … a microarray can contain thousands or tens of thousands relatively short nucleotides of known sequences. ICCBR '10
The end product of a comparative hybridization experiment is a scanned array image. ICCBR '10 Bioinformatics Challenges
ICCBR '10 Bioinformatics Challenges
Microarray applications Determine relative DNA levels associated with huge number of known and predicted genes in a single experiment. The most attractive application of microarrays is in the study of differential gene expression in disease. The up– or down-regulation of gene activity can either be the cause of the pathophysiology or the result of the disease. Accurate measurement of every single gene is assessed. Sensitivity: very high – detect the presence of one transcript in one-tenth of a cell. ICCBR '10 Bioinformatics Challenges
ICCBR '10 Data mining challenges Volume of data (Giga bytes, number of features) Characteristics of data (specific constraints) Domain specific knowledge (expert interpretation) Bioinformatics Challenges
BMA-CBR System ICCBR '10 Gene Expression Level Dataset Application of Feature Selection Algorithm Discrete Sample Output: Supervised Machine Learning and Model Construction through Classification Diagnosis Continuous Sample Output: Supervised Machine Lerning and Model Construction through Prediction Survival analysis
ICCBR '10 BMA-CBR System BMA-CBR system performs feature selection through BMA before using CBR for microarray data classification and prediction (survival analysis) Introduction and motivation of variable selection What is Bayesian Model Averaging (BMA)? One approach: the iterative BMA algorithm Application 1: Chronic Myeloid Leukemia (CML) Application 2: Survival analysis Presentation of CBR
ICCBR '10 Feature selection Used to select a subset of relevant features for building robust learning models in machine learning. Often used in supervised learning. Select relevant features from the training set (for which class labels are known ). Apply the selected features in the test set. Bayesian Model Averaging
ICCBR '10 Feature selection A minimal set of relevant genes for future prediction or assay development Bayesian Model Averaging
ICCBR '10 Typical variable selection methods – one variable at a time Examples: T-test Between group to within group sum of squares (BSS/ WSS) [Dudoit et al. 2001] Bayesian Model Averaging
ICCBR '10 Multivariate gene selection Our goal: consider multiple genes Simultaneously to exploit the interdependence between genes to reduce # relevant genes Bayesian Model Averaging
ICCBR '10 Bayesian Model Averaging (BMA) [Raftery 1995], [Hoeting et. al. 1999] A multivariate variable selection technique. Typical model selection approaches select a model and then proceed as if the selected model has generated the data --> overconfident inferences Advantages of BMA: Fewer selected genes Can be generalized to any number of classes Posterior probabilities for selected genes and selected models Bayesian Model Averaging
ICCBR '10 BMA Average over predictions from several models What do we need? Prediction with a given model k --> logistic regression How to choose a set of “good” models? --> variable selection Bayesian Model Averaging
ICCBR '10 What models to average over? All possible models --> way too many!! Eg. 2^30~1 billion, 2^50~10^15 etc… The BMA solution: 1. “leaps and bounds ” [Furnival and Wilson 1974] : when #variables (genes) <= 30, we can efficiently produce a reduced set of good models (branch and bound). 2. Cut down the # models? Discard models that are much less likely than the best model. Bayesian Model Averaging
ICCBR '10 Iterative BMA algorithm [Yeung, Bumgarner, Raftery 2005] Pre-processing step: Rank genes using BSS/WSS ratio. Initial step: Repeat until all genes are processed: Output: selected genes and models with their posterior probabilities Bayesian Model Averaging
ICCBR '10 Application 1: Classification of progression of chronic myeloid leukemia (CML) Motivation: New Candidates for Prognostic studies in CML Bayesian Model Averaging
ICCBR '10 Progression of CML CML usually presents in chronic phase (CP), but in the absence of effective therapy, CP CML invariably transforms to accelerated phase (AP) disease, and then to an acute leukemia, blast crisis (BC). BC is highly resistant to treatment, and all treatments are more successful when administered during CP. Imatinib is most effective in early CP patients with excellent survival (86% at 7 years). Currently there are limited clinical markers and no molecular tests that can predict the “clock” of CML progression for individual patients at the time of CP diagnosis, making it difficult to adapt therapy to the risk level of each patient. Bayesian Model Averaging
ICCBR '10 Why predictors for CML progression? Bayesian Model Averaging
ICCBR '10 Identification of CML progression biomarkers Bayesian Model Averaging
ICCBR '10 Genes associated with CML progression Bayesian Model Averaging
ICCBR '10 BMA selected genes using microarray data Selected 6 genes over 21 models Repeat CV 100 times Average Brier Score = 0.21 Average prediction accuracy = 99.17% Bayesian Model Averaging
ICCBR '10 PCR data: CP-early vs CP-late Bayesian Model Averaging
ICCBR '10 Summary: CML data BMA applied to a microarray data consisting of patient samples in different phases of CML identified 6 signature genes (ART4, DDX47, IGSF2,LTB4R, SCARB1, SLC25A3). Results validated the gene signature using quantitative PCR: 6- gene signature is highly predictive of CP-early vs CP-late. What is next? To identify biologically meaningful biomarkers for CML progression and response to therapy. Biomarkers that are functionally related (connected in an underlying network) to known reference genes. Bayesian Model Averaging
ICCBR '10 Application 2: Survival analysis Bayesian Model Averaging
ICCBR '10 Results: Breast cancer data Bayesian Model Averaging
ICCBR '10 Results: Breast cancer data - Annest, Bumgarner, Raftery, Yeung. BMC Bioinformatics 2009 Bayesian Model Averaging
CBR Classification task Similarity measure Weights provided by BMA for selected features ICCBR '10
CBR Classification task Choose the class for which the average similar score is highest ICCBR '10
CBR Survival analysis task Similarity measure Weights provided by BMA for selected features ICCBR '10
CBR Survival analysis task Choose the class for which the average similar score is highest ICCBR '10
Evaluation / Classification ICCBR '10 DatasetTotal Number of Samples # Training Samples # Validation Samples Number of Genes Leukemia Leukemia Dataset# classesBMA-CBRiterativeBMAOther published results Leukemia 22#genes = 20 #errors = 1/34 #genes = 20 #errors = 2/34 #genes = 5 #errors = 1/34 Leukemia 33#genes = 15 #errors = 1/34 #genes = 15 #errors = 1/34 #genes ~ 40 #errors = 1/34
Evaluation / Prediction ICCBR '10 DatasetTotal Number# Training Samples # Validation Samples Number Of Genes DLBCL ,399 Breast Cancer ,919 DatasetBMA-CBRiterativeBMABest Other Published Results DLBCL#genes = 25 p-value = #genes = 25 p-value = #genes = 17 p-value = Breast cancer#genes = 15 p-value = 2.14e-10 #genes = 15 p-value = 3.38e-10 #genes = 5 p-value = 3.12e-05
Conclusion The combination of BMA and CBR provides excellent classification and prediction results. It provides promising results for the application of CBR to bioinformatics tasks and data. ICCBR '10
Conclusion Future developments Refine risk classes into more than two risk groups. Refine CBR algorithm. Test on additional datasets. Provide automatic interpretation of the classification / prediction both for gene selection and for case-based reasoning. ICCBR '10