Notes on Biological Networks - Advanced Topics in Programming Language | CMSC 838, Study notes of Computer Science

Download Notes on Biological Networks - Advanced Topics in Programming Language | CMSC 838 and more Study notes Computer Science in PDF only on Docsity! 1 CMSC 838T – Lecture 11 CMSC 838T – Lecture 11 Biological networks 0 Gene networks 0 Gene regulation networks 0 Metabolic networks DNA microarrays 0 Construction 0 Data analysis Affymetrix GeneChip Scanner 3000 CMSC 838T – Lecture 11 Gene Expression Gene expression 0 Genes are expressed when they are transcribed onto RNA 0 Amount of mRNA indicates gene activity No mRNA → gene is off mRNA present → gene is on & performing function Biologically 0 Some genes are always expressed in all tissues Estimated 10,000 housekeeping / ubiquitous genes 0 Other genes are selectively on Depending on tissue, disease, and/or environment 0 Change in environment → change in gene expression So organism can respond 2 CMSC 838T – Lecture 11 Biological Networks Gene expression does not happen in isolation 0 Individual genes code for function Produce mRNA → protein performing function 0 Sets of genes can form pathways Gene products can turn on / off other genes 0 Sets of pathways can form networks When pathways interact Biology is a study of networks 0 Genes 0 Proteins 0 Etc… CMSC 838T – Lecture 11 Biological Networks & DNA Microarrays Overview 0 Biological networks 0 DNA microarray construction 0 Microarray data analysis 5 CMSC 838T – Lecture 11 Examining Biological Networks Indirect approach 0 Measure mRNA production (gene expression) in cell Random ESTs DNA microarray 0 Advantages High throughput Can test large variety of mRNA simultaneously 0 Disadvantages RNA level not always correlated with protein level / function Misses changes at protein level Results may thus be less precise CMSC 838T – Lecture 11 Examining Biological Networks Direct approach 0 Measure protein production / interaction in cell 2D electrophoresis Mass spectroscopy Protein microarray 0 Advantages Precise results on proteins 0 Disadvantages Low throughput (for now) 6 CMSC 838T – Lecture 11 Biological Networks & DNA Microarrays Overview 0 Biological networks 0 DNA microarray construction 0 Microarray data analysis CMSC 838T – Lecture 11 DNA Microarray – Affymetrix System Complete Affymetrix GeneChip instrumentation system 7 CMSC 838T – Lecture 11 DNA Microarray Experimental method for measuring RNA in cell Microarray construction 0 Short single-stranded DNA sequences (probes) cDNA sequences (200+ nucleotides) Oligomers (25-80 nucleotides) 0 Probes attached to glass slide at known fixed locations High precision robotics (spotted cDNA / oligomers) Photolithography (in situ oligomers) 0 Miniaturization is key Measure many (100,000+) genes at once Small amounts mRNA needed Works by hybridization of complementary DNA CMSC 838T – Lecture 11 DNA Microarray – Hybridization Denaturation Separating DNA into single strands Hybridization Forming double-stranded DNA (only if strands are complementary) Heat Cool 10 CMSC 838T – Lecture 11 DNA Microarray – Comparative Hybridization Goal 0 Measure relative amount of mRNA expressed Algorithm 1. Choose cell populations 2. mRNA extraction and reverse transcription 3. Fluorescent labeling of cDNA’s (normalized) 4. Hybridization to microarray 5. Scan the hybridized array 6. Interpret scanned image CMSC 838T – Lecture 11 DNA Microarray – Comparative Hybridization 11 CMSC 838T – Lecture 11 Comparative Hybridization – Output Color determined by relative RNA concentrations Brightness determined by total concentration Gene expressed in A Gene expressed in B Gene expressed in A & B CMSC 838T – Lecture 11 Comparative Hybridization – Issues Choosing cell populations 0 Find cells with selective gene expression 0 Provides hints of gene function Reverse transcription 0 Extract mRNA from cells, purify, transcribe to cDNA 0 mRNA may be partially transcribed, selectively transcribed 0 Result = reverse transcription bias Fluorescent labeling 0 cDNA bound with fluorescent dyes (fluors) 0 Solutions diluted to normalize brightness 0 Assumes fluorescence level directly proportional to mRNA level 12 CMSC 838T – Lecture 11 DNA Microarray – Affymetrix Arrays Construction 0 Synthesize oligomers in situ using photolithography 0 $500,000 per set of masks, $300 per chip Probe set 0 Create multiple oligomers per cDNA Since short individual 25-mers 0 Place negative control next to each probe With exactly one mismatched base at center to track / calibrate mismatches Use 0 Label cDNA, fragment & hybridize 0 Stain labeled cDNA with (single) fluorescent dye 0 Measure intensity using special CCD scanner CMSC 838T – Lecture 11 DNA Microarray – Photolithography Affymetrix DNA microarray 500,000 oligomers in 1.28 cm22) Create 25-mer oligomers on glass slide directly 1) Use photolithography 15 CMSC 838T – Lecture 11 DNA Microarray – Affymetrix Genome Arrays CMSC 838T – Lecture 11 DNA Microarray – Variability & Errors Sources of (undesirable) variability 0 RNA extraction 0 Probe labeling 0 Hybridization kinetics (temperature, time, mixing…) 0 Image analysis 0 Biological variability Sources of error 0 Image artifacts Dust / bubbles in array Spillover from bright spot to neighboring dark spots 0 Self / cross hybridization cDnA hybridize with each other, mismatched probes 16 CMSC 838T – Lecture 11 DNA Microarray – Image Processing Approach 1. Scan the array 2. Quantify each spot 3. Subtract background 4. Normalize intensity (across samples) 5. Calculate expression ratios (log scale) vs. control 6. Export table of fluorescent intensities for each gene Affymetrix software 0 Automatic image processing 0 Precision Around 2% variation in measurements Less than normal biological variability CMSC 838T – Lecture 11 Microarray Image Processing – Expression Ratio Ratio of signal intensity Cy3 signal (log2) C y5 s ig na l ( lo g 2 ) Ratio < –2x Ratio > +2x Calculating expression ratios 0 After filtering, correction, & normalization 0 Find genes with large contrasts in expression level 0 Provides data for single microarray 17 CMSC 838T – Lecture 11 Biological Networks & DNA Microarrays Overview 0 Biological networks 0 DNA microarray construction 0 Microarray data analysis CMSC 838T – Lecture 11 DNA Microarray – Experiment Data Experiment design 0 Measure level of multiple mRNA (i.e., single microarray test) 0 As one or more experimental conditions vary Time elapsed Pathogen / drug exposure Different tissues 0 Result is a multidimensional data mRNA level × tissue × drug exposure × time × … Types of questions 0 What genes are up / down regulated? 0 What genes are over / under-expressed in diseased state? 0 What gene regulation networks exist? 0 Need rigorous statistical analysis to determine significance! 20 CMSC 838T – Lecture 11 DNA Microarray – Pairwise Distances Clustering method may require calculating distance Metric distances 0 Satisfies 4 conditions (for all x,y,z) Positive definite → d(x,y) ≥ 0 Symmetric → d(x,y) = d(y,x) Zero distance to self → d(x,x) = 0 Triangle inequality → d(x,y) ≤ d(x,z) + d(y,z) 0 Example – Euclidean distance Semi-metric distance 0 Satisfies first 3 conditions only (not triangle inequality) 0 Example – Pearson correlation coefficient CMSC 838T – Lecture 11 DNA Microarray – Cluster Distances Merging clusters by minimizing… 0 Inter-cluster distances (single linkage) 0 Maximum intra-cluster distance (complete linkage) 0 Average intra-cluster distances (UPGMA) 0 Distance between center of clusters (centroid) Choice depends on desired efficiency & robustness 0 Single linkage less robust single linkage complete linkage UPGMA / centroid 21 CMSC 838T – Lecture 11 DNA Microarray – Hierarchical Clustering Approach 0 Bottom-up approach (agglomerative) Begin with all genes in individual cluster Repeated merge closest clusters 0 Top-down approach (divisive) Begin with all genes in same cluster Repeatedly split cluster into parts 0 Produces dendogram (unrooted tree) Tim e Genes CMSC 838T – Lecture 11 Microarray – Iterative Clustering Methods K-means clustering 1. Pick K vectors 2. Assign genes to closest of K vectors 3. Pick new K vectors as center of each cluster 4. Repeat until clusters are stable Self-organizing maps (SOM) 1. Pick K partitions 2. User defines geometric configuration for partitions 3. Generate random vector for each partition 4. Randomly pick gene 5. Adjust closest vector to be more similar to vector for gene 6. Repeat until vectors are stable 22 CMSC 838T – Lecture 11 DNA Microarray – SOMs from GeneCluster CMSC 838T – Lecture 11 DNA Microarray – Multivariate Analysis Principal component analysis 0 Linear method 0 Treat every gene as a dimension (vector) 0 Separate genes using singular value decomposition (SVD) Finds linear combinations of vectors to separate data Diagonalization of covariance matrix 0 Projects complex data sets onto reduced dimensionality space 0 Easier to pick out clusters (for use with K-means, SOM) Support vector machine 0 Supervised learning approach 0 Start with positive / negative examples (training set) 0 Train machine to recognize cluster types 0 Use machine to cluster data