Computational Genomics. High-throughput experimental biology

Transcription

1 Computational Genomics /02 810/02-710, Spring 2009 Gene Expression Analysis Data pre-processing processing Eric Xing Lecture 15, March 4, 2009 Reading: class assignment Eric CMU, High-throughput experimental biology Parallel approach to collection of very large amounts of data (by biological standards) Sophisticated instrumentation, requires some understanding Systematic features of the data are at least as important as the random ones Often more like industrial process than single investigator lab research Integration of many data types: clinical, genetic, molecular..databases Eric CMU,

2 Gene Expression Transcription DNA GTAATCCTC CATTAGGAG RNA polymerase mrna GU U AA CC Eric CMU, Measuring Gene Expression Idea: measure the amount of mrna to see which genes are being expressed in (used by) the cell. Measuring protein might be better, but is currently harder. Eric CMU,

3 Reverse transcription Clone cdna strands, complementary to the mrna mrna G U A A U C C U C Reverse transcriptase cdna T T A G G A G C A T T A G G A G C A T T A G G A G C C A T A G G A G C A T C C A A A G T T A G A A G A A G C A A T T T A A G G G G A A G G G C A T T A G G A G Eric CMU, cdna microarray experiments mrna levels compared in many different contexts --- Manifolds in a high-dimensional space spanned by mrna levels Different tissues, same organism (brain v. liver) Same tissue, same organism (ttt v. ctl, tumor v. non-tumor) Same tissue, different organisms (wt v. ko, tg, or mutant) Time course experiments (effect of ttt, development) Other special designs (e.g. to detect spatial patterns). Eric CMU,

4 cdna microarrays cdna clones Eric CMU, Compare the genetic expression in two samples of cells PRINT cdna from one gene on each spot SAMPLES cdna labelled red/green e.g. treatment / control normal / tumor tissue Eric CMU,

5 Imaging HYBRIDIZE Add equal amounts of labelled cdna samples to microarray. Laser SCAN Detector White channel image Eric CMU, Biological question Differentially expressed genes Sample class prediction etc. Experimental design Estimation Microarray experiment 16-bit TIFF files Image analysis (Rfg, Rbg), (Gfg, Gbg) Normalization R, G Testing Clustering Discrimination Biological verification and interpretation Eric CMU,

6 Some statistical questions Image analysis: addressing, segmenting, quantifying Normalisation: within and between slides Quality: of images, of spots, of (log) ratios Which genes are (relatively) up/down regulated? Assigning p-values to tests/confidence to results. Eric CMU, Some statistical and machine learning questions, ctd Planning of experiments: design, sample size Discrimination and allocation of samples Clustering, classification: of samples, of genes Selection of genes relevant to any given analysis Analysis of time course, factorial and other special experiments.....& much more. Eric CMU,

7 Some bioinformatic questions Connecting spots to databases, e.g. to sequence, structure, and pathway databases Discovering short sequences regulating sets of genes: direct and inverse methods Relating expression profiles to structure and function, e.g. protein localisation Identifying novel biochemical or signalling pathways,..and much more. Eric CMU, Image analysis Part of the image of one channel false-coloured on a white (v. high) red (high) through yellow and green (medium) to blue (low) and black scale White channel image Eric CMU,

8 Does one size fit all? Eric CMU, Segmentation: limitation of the fixed circle method Adaptive fit Fixed Circle Inside the boundary is spot (foreground), outside is not. Eric CMU,

9 Quantification of expression For each spot on the slide we calculate Red intensity = Rfg - Rbg fg = foreground, bg = background, and Green intensity = Gfg - Gbg and combine them in the log (base 2) ratio Log2( Red intensity / Green intensity) Eric CMU, Gene Expression Data On p genes for n slides: p is O(10,000), n is O(10-100), but growing, Slides Genes slide 1 slide 2 slide 3 slide 4 slide Gene expression level of gene 5 in slide 4 = Log 2 ( Red intensity / Green intensity) These values are conventionally displayed on a red (>0) yellow (0) green (<0) scale. Eric CMU,

10 Eric CMU, Systematic errors Probes: ~6,000 cdnas, including 200 related to lipid metabolism. Arranged in a 4x4 array of 19x21 sub-arrays. Eric CMU,

11 Analysis of possible systematic errors Data pre-processing Eric CMU, Quality of the signals Signal/Noise = log 2 (spot intensity/background intensity) Eric CMU,

12 Spatial plots: background from the two slides Eric CMU, The red/green ratios can be spatially biased Top 2.5%of ratios red, bottom 2.5% of ratios green Eric CMU,

13 The red/green ratios can be intensity-biased Eric CMU, Intensity Bias (within array) log 2 R vs log 2 G M=log 2 R/G vs A=log 2 RG Eric CMU,

14 Intensity Bias (across experiments) Boxplots of log 2 R/G Liver samples from 16 mice: 8 WT, 8 ApoAI KO. Eric CMU, Plate effects Eric CMU,

15 Spatial Bias Highlighting extreme log ratios Top (black) and bottom (green) 5% of log ratios Eric CMU, Print-tip group effects Log-ratios Print-tip groups Clear example of spatial bias Eric CMU,

16 Pin group (sub-array) effects Lowess lines through points from pin groups Boxplots of log ratios by pin group Eric CMU, Time of printing effects spot number Green channel intensities (log 2 G). Printing over 4.5 days. The previous slide depicts a slide from this print run. Eric CMU,

17 Normalization Why? To correct for systematic differences between samples on the same slide, or between slides, which do not represent true biological variation between samples. How do we know it is necessary? By examining self-self hybridizations, where no true differential expression is occurring. We find dye biases which vary with overall spot intensity, location on the array, plate origin, pins, scanning parameters,. Eric CMU, Self-self comparison False color overlay Boxplots within pin-groups Scatter (MA-)plots Eric CMU,

18 A series of non self-self comparison From the NCI60 data set (Stanford web site) Eric CMU, Early Ngai lab, UC Berkeley Eric CMU,

19 Early Goodman lab, UC Berkeley Eric CMU, From the Ernest Gallo Clinic & Research Center Eric CMU,

20 Early PMCRI, Melbourne Australia Eric CMU, Normalization: methods a) Normalization based on a global adjustment log 2 R/G -> log 2 R/G - c = log 2 R/(kG) Choices for k or c = log 2 k are c = median or mean of log ratios for a particular gene set (e.g. housekeeping genes). Or, total intensity normalization, where k = R i / G i. b) Intensity-dependent normalization. Here we run a line through the middle of the MA plot, shifting the M value of the pair (A,M) by c=c(a), i.e. log 2 R/G -> log 2 R/G - c (A) = log 2 R/(k(A)G). One estimate of c(a) is made using the LOWESS function of Cleveland (1979): LOcally WEighted Scatterplot Smoothing. Eric CMU,

21 Normalization: methods c) Within print-tip group normalization. In addition to intensity-dependent variation in log ratios, spatial bias can also be a significant source of systematic error. Most normalization methods do not correct for spatial effects produced by hybridization artefacts or print-tip or plate effects during the construction of the microarrays. It is possible to correct for both print-tip and intensity-dependent bias by performing LOWESS fits to the data within print-tip groups, i.e. log 2 R/G -> log 2 R/G - c i (A) = log 2 R/(k i (A)G), where c i (A) is the LOWESS fit to the MA-plot for the ith grid only. Eric CMU, Which spots to use for normalization? The LOWESS lines can be run through many different sets of points, and each strategy has its own implicit set of assumptions justifying its applicability. For example, we can justify the use of a global LOWESS approach by supposing that, when stratified by mrna abundance, a) only a minority of genes are expected to be differentially expressed, or b) any differential expression is as likely to be up-regulation as down-regulation. Pin-group LOWESS requires stronger assumptions: that one of the above applies within each pin-group The use of other sets of genes, e.g. control or housekeeping genes, involve similar assumptions. Eric CMU,

22 Use of control spots M = log R/G = logr - logg A = ( logr + logg) /2 Eric CMU, Global scale global lowess pin-group lowess spatial plot after Eric CMU,

23 MSP titration series (Microarray Sample Pool) Pool the whole library Control set to aid intensity- dependent normalization Different concentrations Spotted evenly spread across the slide Eric CMU, MSP normalization compared to other methods Orange: Schadt-Wong rank invariant set, Red line: lowess smooth Yellow: GAPDH, tubulin Light blue: MSP pool / titration Eric CMU,

24 Composite normalization c i (A)=α A g(a)+(1-α A )f i (A) Before and after composite normalization -MSP lowess curve -Global lowess curve -Composite lowess curve (Other colours control spots) Eric CMU, Comparison of Normalization Schemes No consensus on best segmentation or normalization method Scheme was applied to assess the common normalization methods Based on reciprocal labeling experiment data for a series of 140 replicate experiments on two different arrays each with 19,200 spots Eric CMU,

25 DESIGN OF RECIPROCAL LABELING EXPERIMENT Replicate experiment in which we assess the same mrna pools but invert the fluors used. The replicates are independent experiments and are scanned, quantified and normalized as usual Eric CMU, RECIPROCAL LABELING EXPERIMENT The following relationship would be observed for reciprocal microarray experiments in which the slides are free of defects and the normalization scheme performed ideally log Ch1/ Ch2 Ch1/ Ch2 2 ( RatioGeneA ) Exp.1 = log2( RatioGeneA ) Exp.2 We can measure using real data sets how well each microarray normalization scheme approaches this ideal Eric CMU,

26 Deviation metric to assess normalization schemes Deviation = log + Spot Ch1/ Ch2 Ch1/ Ch2 2 ( RatioGeneA ) Exp.1 log2( RatioGeneA ) Exp.2 Deviation ArrayAverage = n 1 Ch1/ Ch2 log ( Ratio ) + log 2 GeneN Exp.1 n 2 ( Ratio Ch1/ Ch2 GeneN ) Exp.2 We now use the mean array average deviation to compare the normalization methods. Note that this comparison addresses only variance (precision) and not bias (accuracy) aspects of normalization Eric CMU, Comparison of Normalization Schemes Comparison of Normalization Methods - Using K Microarrays Average Mean Deviation Value *** Pre Normalized Global Intensity Subarray Intensity Global Ratio Sub-Array Ratio Global LOWESS Subarray LOWESS Normalization Method Eric CMU,

27 Scale normalization: between slides Boxplots of log ratios from 3 replicate self-self hybridizations. Left panel: before normalization Middle panel: after within print-tip group normalization Right panel: after a further between-slide scale normalization. Eric CMU, The NCI 60 experiments (no bg) Some scale normalization seems desirable Eric CMU,

28 One way of taking scale into account Assumption: All slides have the same spread in M True log ratio is µ ij where i represents different slides and j represent different spots. Observed is M ij, where M ij = a i µ ij Robust estimate of a i is I MAD i I i =1 MAD i MAD i = median j { y ij - median(y ij ) } Eric CMU, A slightly harder normalization problem Global lowess doesn t do the trick here. Eric CMU,

29 Print-tip-group normalization helps Eric CMU, But not completely There is still a lot of scatter in the middle in a WT vs KO comparison. Eric CMU,

30 Effects of previous normalization Before normalisation After print-tip-group normalization Eric CMU, Within print-tip-group box plots of M after print-tip-group normalization Eric CMU,

31 Taking scale into account, cont. Assumption: All print-tip-groups have the same spread in M True log ratio is µ ij where i represents different print-tip-groups and j represent different spots. Observed is M ij, where M ij = a i µ ij Robust estimate of a i is MAD i = median j { y ij - median(y ij ) } I MAD i I i =1 MAD i Eric CMU, Effect of location & scale normalization Clearly care is needed in making decisions like this one. Eric CMU,

32 A comparison of three M v A plots Unnormalized Print-tip normalization Print tip & scale n. Eric CMU, Follow-up experiment On each slide, half the spots ( 8) are differentially expressed, the other half are not. Eric CMU,

33 Paired-slides: dye-swap Slide 1, M = log 2 (R/G) - c Slide 2, M = log 2 (R /G ) - c Combine by subtracting the normalized log-ratios: [ (log 2 (R/G) - c) - (log 2 (R /G ) - c ) ] / 2 [ log 2 (R/G) + log 2 (G /R ) ] / 2 [ log 2 (RG /GR ) ] / 2 provided c = c. Assumption: the normalization functions are the same for the two slides. Eric CMU, Checking the assumption MA plot for slides 1 and 2: it isn t always like this. Eric CMU,

34 Result of self-normalization (M - M )/2 vs. (A + A )/2 Eric CMU, Summary of normalization Reduces systematic (not random) effects Makes it possible to compare several arrays Use logratios (M vs A-plots) Lowess normalization (dye bias) MSP titration series composite normalization Pin-group location normalization Pin-group scale normalization Between slide scale normalization More? Use controls! Normalization introduces more variability Outliers (bad spots) are handled with replication Eric CMU,