Microarray Data Analysis Using BASE Danny Park MGH Microarray Core March 15, 2004

You’ve got data! What was I asking? – remember your experimental design How do I analyze the data? –How do I find interesting stuff? – learn some analysis tools –How do I trust the results? – statistics is key

What was I asking? Typically: “which genes changed expression levels when I did ____” Common ____: –Binary conditions: knock out, treatment, etc –Continuous scales: time courses, levels of treatment, etc –Unordered discrete scales: multiple types of treatment or mutations This tutorial’s focus: binary experiments

How do I analyze the data? BASE – BioArray Software Environment –Data storage and distribution –Simple filtering, normalization, averaging, and statistics –Export/Download results to other tools MS Excel TIGR Multi Experiment Viewer (TMEV) This tutorial’s focus: using BASE

Today’s Presentation Demonstrate the most basic analysis techniques Using our most frequently used software (BASE) For the most common kind of experiments

Work Flow Images & data files scan, segment upload BASE Labeled cDNA Slides QC & label hybridize RNA analysis Researcher

The Most Common experiment Two-sample comparison w/N replicates –KO vs. WT –Treated vs. untreated –Diseased vs. normal –Etc Question of interest: which genes are (most) differentially expressed?

Experimental Design – naïve A B From Gary Churchill, Jackson Labs

Experimental Design – tech repl A B From Gary Churchill, Jackson Labs

Experimental Design – bio repl  Treatment  Biological Replicate  Technical Replicate  Dye  Array ABA B From Gary Churchill, Jackson Labs

The Most Common Analysis Filter out bad spots Adjust low intensities Normalize – correct for non-linearities and dye inconsistencies Filter out dim spots Calculate average fold ratios and p- values per gene Rank, sort, filter, squint, sift data Export to other software

MGH BASE is a microarray data storage and analysis package BASE resides on our web server –Data is stored at our facility –Computation is performed on our machines All you need is a web browser – –A Microarray Core technician will provide you with a username, password, and experiment name

BASE – Login page

BASE – Logged in

BASE – Sidebar Reporters

BASE – Sidebar Reporters

BASE – Sidebar Array LIMS

BASE – Sidebar Array LIMS

BASE – Sidebar Biomaterials

BASE – Sidebar Biomaterials

BASE – Sidebar Hybridizations

BASE – Sidebar Hybridizations

BASE – Sidebar Analyze Data

BASE – Sidebar Analyze Data

BASE – Sidebar Users

BASE – Sidebar Users

BASE – My Account Change your password and access defaults

BASE – My Account Change your password and access defaults

BASE – My Account Change your password and access defaults

BASE – My Account Change your password and access defaults

Find your experiment

Experiment view: Four Tabs

Group slide data together

Select the slides that measure the same thing. Later in analysis, they will be averaged together. In this experiment, all ten slides are replicates, so there is only one grouping.

Group slide data together Select the slides that measure the same thing. Later in analysis, they will be averaged together. In this experiment, all ten slides are replicates, so there is only one grouping.

Group slide data together Select the slides that measure the same thing. Later in analysis, they will be averaged together. In this experiment, all ten slides are replicates, so there is only one grouping.

Group slide data together

Give your data set a descriptive name to distinguish it from other slide groupings. In this Myd88 knockout experiment, there is only one grouping, so a generic name is fine.

Group slide data together Give your data set a descriptive name to distinguish it from other slide groupings. In this Myd88 knockout experiment, there is only one grouping, so a generic name is fine.

Group slide data together Give your data set a descriptive name to distinguish it from other slide groupings. In this Myd88 knockout experiment, there is only one grouping, so a generic name is fine.

Analysis: Begin

Analysis: Filter Setup “Bad” spots are marked with a negative Flag value.

Analysis: Filter Setup “Bad” spots are marked with a negative Flag value.

Analysis: Filter Setup “Bad” spots are marked with a negative Flag value.

Analysis: Filter Setup “Bad” spots are marked with a negative Flag value.

Analysis: Filter Setup “Bad” spots are marked with a negative Flag value.

Analysis: Filter Setup “Bad” spots are marked with a negative Flag value.

Analysis: Filter Setup “Bad” spots are marked with a negative Flag value.

Analysis: Filter Setup “Bad” spots are marked with a negative Flag value.

Analysis: Filter Setup “Bad” spots are marked with a negative Flag value. Oligos are annotated with species codes, but control spots are not. Set species to your two-letter code of choice (Mm, Hs, Dr, Pa, etc)

Analysis: Filter Setup “Bad” spots are marked with a negative Flag value. Oligos are annotated with species codes, but control spots are not. Set species to your two-letter code of choice (Mm, Hs, Dr, Pa, etc)

Analysis: Filter Setup “Bad” spots are marked with a negative Flag value. Oligos are annotated with species codes, but control spots are not. Set species to your two-letter code of choice (Mm, Hs, Dr, Pa, etc)

Analysis: Filter Setup “Bad” spots are marked with a negative Flag value. Oligos are annotated with species codes, but control spots are not. Set species to your two-letter code of choice (Mm, Hs, Dr, Pa, etc)

Analysis: Filter Setup “Bad” spots are marked with a negative Flag value. Oligos are annotated with species codes, but control spots are not. Set species to your two-letter code of choice (Mm, Hs, Dr, Pa, etc)

Analysis: Filter Setup “Bad” spots are marked with a negative Flag value. Oligos are annotated with species codes, but control spots are not. Set species to your two-letter code of choice (Mm, Hs, Dr, Pa, etc)

Analysis: Filter Setup “Bad” spots are marked with a negative Flag value. Oligos are annotated with species codes, but control spots are not. Set species to your two-letter code of choice (Mm, Hs, Dr, Pa, etc)

Analysis: Filter Setup “Bad” spots are marked with a negative Flag value. Oligos are annotated with species codes, but control spots are not. Set species to your two-letter code of choice (Mm, Hs, Dr, Pa, etc)

Analysis: Filter Setup Naming the filter and the child data set are essential to reducing confusion later.

Analysis: Filter Setup Naming the filter and the child data set are essential to reducing confusion later.

Analysis: Filter Setup Naming the filter and the child data set are essential to reducing confusion later.

Analysis: Filter Run

Analysis: Quality Data

Analysis: Unfiltered Data

Analysis: Filter Parameters

Analysis: Limit-Int Setup

Analysis: Check job status

“All done” indicates the job is complete.

Analysis: Check job status “All done” indicates the job is complete.

Analysis: Limit-Int Output

Analysis: Change data set name

Change the name of this set to “Intensity limited Data”

Analysis: Change data set name

Analysis: LOWESS Setup

Analysis: Check job status

Analysis: LOWESS Output

Change the name of this set to “Normalized Data” using the same steps as before.

Analysis: Change data set name Change the name of this set to “Normalized Data” using the same steps as before.

Analysis: Change data set name Change the name of this set to “Normalized Data” using the same steps as before.

Analysis: Filter Setup Set up the filter as indicated, hit Add/Update on the Gene filter, then hit Accept and select the resulting data set.

Analysis: Useful Data

MA Plots: Raw Myd88 Data

MA Plots: Quality Data

MA Plots: Int-limited Data

MA Plots: Normalized Data

MA Plots: Norm. Corr. Factor

MA Plots: Useful Data

Analysis: Useful Data

Analysis: Fold Ratio Setup

Analysis: Fold Ratio Output

Analysis: Change list name

Change the name of this list as indicated here.

Analysis: Change list name Change the name of this list as indicated here.

Analysis: Change list name

Analysis: Fold Ratio Graphs

Analysis: t-test Setup

Analysis: t-test Output

Analysis: Change list name Change the name of this set to “myd88 p- value” using the same steps as before.

Analysis: Change list name Change the name of this set to “myd88 p- value” using the same steps as before.

Analysis: Change list name Change the name of this set to “myd88 p- value” using the same steps as before.

Analysis: t-test Graphs

Analysis: Experiment Explorer

EExplore: Single Gene View

EExplore: Gene List View

Fill out the table as indicated, then hit Add/Update.

EExplore: Gene List View

EExplore: NCBI Links

EExplore: Gene List View This additional row will restrict hits to P values of 5% or less.

EExplore: Gene List View This additional row will restrict hits to P values of 5% or less.

EExplore: Single Gene View

EExplore: Gene List View

Open MS Excel and tell it to open the file you downloaded (typically called base.tsv).

EExplore: Gene List View Open MS Excel and tell it to open the file you downloaded (typically called base.tsv).

Have Fun! The rest of the analysis is largely driven by your biological understanding of the genes indicated in these lists. We cannot help much in the interpretation of this data. Don’t forget to go back to the raw data sets and repeat this entire analysis for any other slide groupings.

Acknowledgements MGH Lipid Metabolism Unit Mason Freeman Harry Bjorkbacka MGH Lipid Metabolism Unit Mason Freeman Harry Bjorkbacka LUND (Sweden) Dept. Theoretical Physics & Dept. Oncology Carl Troein Lao H. Saal Johan Vallon-Christersson Sofia Gruvberger Åke Borg Carsten Peterson LUND (Sweden) Dept. Theoretical Physics & Dept. Oncology Carl Troein Lao H. Saal Johan Vallon-Christersson Sofia Gruvberger Åke Borg Carsten Peterson MGH Microarray Core Glenn Short Jocelyn Burke Najib El Messadi Jason Frietas Zhiyong Ren MGH Microarray Core Glenn Short Jocelyn Burke Najib El Messadi Jason Frietas Zhiyong Ren MGH Molecular Biology Bioinformatics Group Chuck Cooper Xiaowei Wang Harvard School of Public Health Biostatistics Xiaoman Li MGH Molecular Biology Bioinformatics Group Chuck Cooper Xiaowei Wang Harvard School of Public Health Biostatistics Xiaoman Li