0.25
0.5
0.75
1.25
1.5
1.75
2
Bioinformatics Challenge: Learning in Very High Dimensions with Very Few Samples
Published on Feb 25, 20077469 Views
Dedicated machine learning procedures have already become an integral part of modern genomics and proteomics. However, these very high dimensional and low learning sample tasks often stretch these pro
Related categories
Chapter list
Learning in Very High Dimensions with<br>Very Few Samples00:00
Part 1: Cancer genomics00:51
Microarray background00:58
Microarray analysis of gene expression06:40
Principles of microarray10:13
Reproducibility of array experiments10:58
Data normalization12:06
PCR amplification13:18
PCR amplification14:17
Low density Q-PCR Array (ABI)14:43
Peter MacCallum<br> Cancer Centre15:29
Carcinomas of Unknown Primary16:12
Cancer in brief16:58
Data collection18:04
Tumour samples of<br>known origin19:15
Developing a training set of expression profiles19:43
Predictive modeling for gene<br>expressions data21:08
Classifier23:42
Supervised Leaning <br>Support Vector Machine (SVM) – Binary Classifier24:57
Decision margin25:31
Summary of cross-validation<br>test26:03
LOO CV Confusion Table for<br> combined classifier28:12
Test on metastases29:32
Cross platform translation: <br>form microarrays to low-density Q-PCR31:26
Q-PCR site of origin diagnostic - Pilot Study32:25
Comparison of cDNA microarray<br>and Q-PCR data32:57
Data sets34:55
Data transformations for<br>cross platform model transfer35:02
Comparison of 5-class and<br>6-class models36:43
Summary of SVM tests36:49
Comparison of 5- and 6-<br>class models36:54
References38:51
Acknowledgements39:06