The document discusses data integration and summarizes the STITCH database, which integrates protein-small molecule interactions from several sources. It also summarizes a method for predicting novel drug targets using side effect information by analyzing similarities in side effect profiles between existing drugs. Predictions were tested in vitro and in cell assays, with promising results. The document acknowledges contributions from researchers involved in developing STITCH and the side effect prediction method.

1. Data integration The STITCH database of protein–small molecule interactions Lars Juhl Jensen

2. Kuhn et al., Nucleic Acids Research , 2010

3. functional associations

4. protein–small molecule

5. protein–protein

6. parts lists

7. >2.5 million proteins

8. 630 genomes

9. many databases

10. different formats

11. model organism databases

12. Ensembl

13. RefSeq

14. PubChem compounds

15. >74,000 small molecules

16. curated knowledge

17. complexes

18. pathways

19. Letunic & Bork, Trends in Biochemical Sciences , 2008

20. high confidence

21. many databases

22. MIPS Munich Information center for Protein Sequences

23. Gene Ontology

24. KEGG Kyoto Encyclopedia of Genes and Genomes

25. MetaCyc

26. PID NCI-Nature Pathway Interaction Database

27. Reactome

28. different formats

29. different identifiers

30. partially redundant

31. interaction data

32. protein–small molecule

33. in vitro binding assays

34. protein–protein

35. yeast two-hybrid

36. affinity purification

37. fragment complementation

38. Jensen & Bork, Science , 2008

39. genetic interactions

40. Beyer et al., Nature Reviews Genetics , 2007

41. gene coexpression

43. many databases

44. BindingDB

45. CTD Comparative Toxicogenomics Database

46. DrugBank

47. GLIDA GPCR-Ligand Database

48. PDSP K i Psycoactive Drug Screening Program

49. PharmGKB Pharmacogenomics Knowledge Base

50. BIND Biomolecular Interaction Network Database

51. BioGRID General Repository for Interaction Datasets

52. DIP Database of Interacting Proteins

53. IntAct

54. MINT Molecular Interactions Database

55. HPRD Human Protein Reference Database

56. PDB Protein Data Bank

57. GEO Gene Expression Omnibus

58. different formats

59. different identifiers

60. partially redundant

61. literature mining

62. >10 km

63. human readable

64. not computer readable

65. different names

66. text corpus

67. M EDLINE

68. SGD Saccharomyces Genome Database

69. The Interactive Fly

70. OMIM Online Mendelian Inheritance in Man

71. dictionary

72. co-mentioning

73. NLP Natural Language Processing

75. restricted access

76. genomic context

77. gene fusion

78. Korbel et al., Nature Biotechnology , 2004

79. conserved neighborhood

80. operons

81. Korbel et al., Nature Biotechnology , 2004

82. bidirectional promoters

83. Korbel et al., Nature Biotechnology , 2004

84. phylogenetic profiles

85. Korbel et al., Nature Biotechnology , 2004

86. integration

87. many data types

88. not comparable

89. variable quality

90. spread over 630 genomes

91. quality scores

92. reproducibility

93. von Mering et al., Nucleic Acids Research , 2005

94. intergenic distances

95. Korbel et al., Nature Biotechnology , 2004

96. benchmarking

97. calibrate vs. gold standard

98. von Mering et al., Nucleic Acids Research , 2005

99. raw quality scores

100. probabilistic scores

101. orthology transfer

102. von Mering et al., Nucleic Acids Research , 2005

103. combine all evidence

105. Acknowledgments Michael Kuhn Monica Campillos Christian von Mering Manuel Stark Samuel Chaffron Philippe Julien Tobias Doerks Jan Korbel Berend Snel Martijn Huynen Peer Bork

106. Predicting novel targets for existing drugs using side effect information Lars Juhl Jensen

107. the problem

108. new uses for old drugs

109. drug–drug network

110. shared target(s)

111. chemical similarity

112. Campillos & Kuhn et al., Science , 2008

113. Campillos & Kuhn et al., Science , 2008

114. similar drugs share targets

115. only trivial predictions

116. the idea

117. chemical perturbations

118. phenotypic readouts

119. drug treatment

120. side effects

121. the implementation

122. information on side effects

123. package inserts

124. Campillos & Kuhn et al., Science , 2008

125. text mining

126. side-effect ontology

127. backtracking

128. Campillos & Kuhn et al., Science , 2008

129. side-effect correlations

130. Campillos & Kuhn et al., Science , 2008

131. GSC weighting

132. side-effect frequencies

133. Campillos & Kuhn et al., Science , 2008

134. raw similarity score

135. Campillos & Kuhn et al., Science , 2008

136. p-values

137. Campillos & Kuhn et al., Science , 2008

138. side-effect similarity

139. chemical similarity

140. Campillos & Kuhn et al., Science , 2008

141. reference set

142. drug–target pairs

143. Campillos & Kuhn et al., Science , 2008

144. drug–drug pairs

145. score bins

146. benchmark

147. Campillos & Kuhn et al., Science , 2008

148. fit calibration function

149. Campillos & Kuhn et al., Science , 2008

150. probabilistic scores

151. the results

152. drug–drug network

153. ATC codes

154. Campillos & Kuhn et al., Science , 2008

155. categorization

156. Campillos & Kuhn et al., Science , 2008

157. Campillos & Kuhn et al., Science , 2008

158. Campillos & Kuhn et al., Science , 2008

159. map onto score space

160. Campillos & Kuhn et al., Science , 2008

161. the experiments

162. 20 drug–drug relations

163. in vitro binding assays

164. Campillos & Kuhn et al., Science , 2008

165. Campillos & Kuhn et al., Science , 2008

166. Campillos & Kuhn et al., Science , 2008

167. K i <10 µM for 11 of 20

168. cell assays

169. Campillos & Kuhn et al., Science , 2008

170. 9 of 9 showed activity

171. the future

172. SIDER

173. integration with STITCH

174. Acknowledgments Monica Campillos Michael Kuhn Anne-Claude Gavin Peer Bork

175. larsjuhljensen