overview6 hooks

Views:
 
     
 

Presentation Description

No description available.

Comments

Presentation Transcript

Slide1: 

Shooting only Bangor Bioinformatics Mark A. Hooks School of Biological Sciences

Slide2: 

Bioinformatic Interests Zoological/Ecological Tree-fitting Population Structure Analysis Genomics/Post-genomics Genotyping ‘omics’ Health Clinical Image analysis Bioengineering Image analysis

Slide3: 

The SBS Post-Genomics Platform Suite of instrumentation to facilitate investigations of gene function. Capabilities: Transcript profiling through arrays Metabolite profiling by mass spec. Protein profiling by 2D gels and mass spec.

Slide4: 

Metabolomics Proteomics Detection limit = 10-18 moles Mass accuracy < 1 PPM error Bruker Reflex IV MALDI-ToF MS Bruker Apex IV FTICRMS Automated Spot Picking ProEXPRESS Proteomic Imaging System Arthur Multiwavelength fluorescent scanner Investigor ProPic Peptide analysis Transcriptomics Liquid handling & spotting Slide Scanning (Affymetrix 428) Image processing - Imagene Data Analysis - Genesight

Slide5: 

Functional Genomics Investigating relationships between metabolite and transcript levels Gene function Delineating functional pathways How do metabolites influence gene expression? What metabolites influence gene expression? Stress or developmental transitions Global regulators

Slide6: 

If a gene is responsive to a metabolite, then it’s expression will change in relation to that metabolite’s concentration. If a metabolite induces expression there will be a positive correlation coefficient for transcript level vs metabolite level and visa versa. Metabolite Transcript

Slide7: 

Resolve through mutant analysis Potential metabolic signal Explore mechanism

Slide8: 

Grouping metabolites within interacting units Basic visual analysis - Clustered Image Maps Metabolites (clustered)

Slide9: 

Slightly more complex system Gene Metabolite Protein

Slide10: 

Classifiers Control Mutant Unsupervised Supervised Feature(s) Discriminate

Slide11: 

Combining classifiers Classifier ensembles linear quadratic neural networks Parzen logistic support vector machines fuzzy…………… naïve Bayes decision tree nearest neighbour histogram-based Dr. Ludmila Kuncheva (School of Informatics)

Slide12: 

Combiner The input vector The output label Classifier Classifier Classifier . . . x Combining classifiers:

Slide13: 

Methods for building classifier ensembles 1. Optimize the combiner 3. Try different classifier models e.g., decision trees, neural networks, nearest neighbor, etc. 4. Train each classifier on different subset of features 5. Alter the training data: sample from the whole data set, inject noise (bagging, boosting, random forests) Classifier . . . x 2. Use error-correcting output codes (ECOC) Classifier Combiner

Slide14: 

Scientific visualization School of Informatics Prof. Nigel W. John Dr. Thomas Varsamidis Dr. Ik Soo Lim Medical Imaging Biogeometry Human - Computer interfacing

Slide15: 

A Biomedical Knowledge Grid: Towards an Architecture for Data Mining and Knowledge Discovery for Biomedical Resources in a Grid Bill Teahan wjt@informatics.bangor.ac.uk Artificial Intellegence Group Alex Colquhoun HGMD Research group (Cardiff University) Seamless integration of geographically distributed information resources

Slide16: 

Professor Ron Pethig (Aura BioSystems Inc.) Dr. Julian Burt Laboratory on a chip technology Dielectrophoresis for non-invasive cell characterisation of cell properties

authorStream Live Help