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

2. 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

3. 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

4. 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

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

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

7. 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?

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

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

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

11. 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

12. BASE @ 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 –https://base.mgh.harvard.edu/https://base.mgh.harvard.edu/ –A Microarray Core technician will provide you with a username, password, and experiment name

13. BASE – Login page

17. BASE – Logged in

19. BASE – Sidebar Reporters

20. BASE – Sidebar Reporters

21. BASE – Sidebar Array LIMS

22. BASE – Sidebar Array LIMS

23. BASE – Sidebar Biomaterials

24. BASE – Sidebar Biomaterials

25. BASE – Sidebar Hybridizations

26. BASE – Sidebar Hybridizations

27. BASE – Sidebar Analyze Data

28. BASE – Sidebar Analyze Data

29. BASE – Sidebar Users

30. BASE – Sidebar Users

31. BASE – My Account Change your password and access defaults

32. BASE – My Account Change your password and access defaults

33. BASE – My Account Change your password and access defaults

34. BASE – My Account Change your password and access defaults

35. Find your experiment

39. Experiment view: Four Tabs

47. Group slide data together

48. 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.

49. 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.

50. 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.

51. Group slide data together

52. 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.

53. 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.

54. 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.

55. Analysis: Begin

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

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

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

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

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

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

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

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

67. 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)

68. 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)

69. 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)

70. 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)

71. 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)

72. 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)

73. 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)

74. 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)

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

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

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

78. Analysis: Filter Run

79. Analysis: Quality Data

81. Analysis: Unfiltered Data

82. Analysis: Filter Parameters

83. Analysis: Limit-Int Setup

89. Analysis: Check job status

93. “All done” indicates the job is complete.

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

95. Analysis: Limit-Int Output

101. Analysis: Change data set name

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

104. Analysis: Change data set name

108. Analysis: LOWESS Setup

114. Analysis: Check job status

116. Analysis: LOWESS Output

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

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

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

121. 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.

122. Analysis: Useful Data

124. MA Plots: Raw Myd88 Data

128. MA Plots: Quality Data

134. MA Plots: Int-limited Data

140. MA Plots: Normalized Data

146. MA Plots: Norm. Corr. Factor

148. MA Plots: Useful Data

154. Analysis: Useful Data

156. Analysis: Fold Ratio Setup

160. Analysis: Fold Ratio Output

168. Analysis: Change list name

170. Change the name of this list as indicated here.

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

172. Analysis: Change list name

174. Analysis: Fold Ratio Graphs

180. Analysis: t-test Setup

184. Analysis: t-test Output

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

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

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

193. Analysis: t-test Graphs

199. Analysis: Experiment Explorer

201. EExplore: Single Gene View

207. EExplore: Gene List View

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

211. EExplore: Gene List View

219. EExplore: NCBI Links

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

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

222. EExplore: Single Gene View

228. EExplore: Gene List View

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

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

231. 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.

232. 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