Pathway based OMICs data classification
-
Upload
luca-vitale -
Category
Education
-
view
32 -
download
2
Transcript of Pathway based OMICs data classification
![Page 1: Pathway based OMICs data classification](https://reader031.fdocuments.us/reader031/viewer/2022030317/5a6532ee7f8b9a76138b475d/html5/thumbnails/1.jpg)
PathwaybasedOMICsdataclassification
Bioinformatics- 2016/2017
![Page 2: Pathway based OMICs data classification](https://reader031.fdocuments.us/reader031/viewer/2022030317/5a6532ee7f8b9a76138b475d/html5/thumbnails/2.jpg)
Goals
• Classificationwithpathways- Groupofgenesthatareinvolvedinthesamebiologicalfunctions
• Identifyrelationsamongpathways
• BuildagraphofinteractionsbetweenpathwaysandmiRNAs
![Page 3: Pathway based OMICs data classification](https://reader031.fdocuments.us/reader031/viewer/2022030317/5a6532ee7f8b9a76138b475d/html5/thumbnails/3.jpg)
Data
• BreastCancer(BC)• 151patients
• RNA(20501genes)• miRNA(1046)
• 4classes• LumA - 55• LumB - 59• Basal- 24• Her2- 13
PathwaysbyMSigDBà KEGG,Reactome,Biocarta,C6…
• Glioma• 167patients
• RNA(12042genes)• miRNA(534)
• 4classes• Proneural - 52• Classical- 37• Mesenchymal- 54• Neural- 24
![Page 4: Pathway based OMICs data classification](https://reader031.fdocuments.us/reader031/viewer/2022030317/5a6532ee7f8b9a76138b475d/html5/thumbnails/4.jpg)
Introduction
DiscriminantFuzzyPatterns
EnrichmentAnalysis
ClassificationlinearSVM
PermutationTest
Genes
miRNAsMSigDB
InteractionGraph
![Page 5: Pathway based OMICs data classification](https://reader031.fdocuments.us/reader031/viewer/2022030317/5a6532ee7f8b9a76138b475d/html5/thumbnails/5.jpg)
Firststep
• BCDatasetà TrainingSet(70%)andTestSet(30%)
• GliomaDatasetà TrainingSet(75%)andTestSet(25%)
![Page 6: Pathway based OMICs data classification](https://reader031.fdocuments.us/reader031/viewer/2022030317/5a6532ee7f8b9a76138b475d/html5/thumbnails/6.jpg)
Featureselection
• Discriminantfuzzypattern• Toomanyfeaturesà Identifydiscriminantgenes
• Enrichmentà Groupinggenesinpathways(MSigDB)• IdentifywhichpathwaysaresignificantlyrepresentedbythegenesselectedwiththeDFPalgorithm
![Page 7: Pathway based OMICs data classification](https://reader031.fdocuments.us/reader031/viewer/2022030317/5a6532ee7f8b9a76138b475d/html5/thumbnails/7.jpg)
DiscriminantFuzzyPattern– Gridsearch(BC)
• Skipfactorà 0,1,2,3• Factortoskipoutliers.Lowervalueàmorevaluesskipped(0:don’tskip)
• Zetaà 0.35,0.4,0.45,0.5• Threshold used inthemembership functions tolabel thefloatvalues withadiscretevalue
• piValà 0.4,0.45,0.5,0.55,0.6,0.65,0.7,0.75,0.8• Percentage ofvalues ofaclass todetermine thefuzzy patterns
• Overlappingà 1,2• Determines thenumber ofdiscretelabels
• Genes after DFPà 578withSkip Factor 1,Zeta0.4,piVal 0.65andOverlapping 1
![Page 8: Pathway based OMICs data classification](https://reader031.fdocuments.us/reader031/viewer/2022030317/5a6532ee7f8b9a76138b475d/html5/thumbnails/8.jpg)
DiscriminantFuzzyPattern– Gridsearch(Glioma)• Skipfactorà 1,2,3• Zetaà 0.35,0.4,0.45,0.5• piValà 0.6,0.65,0.7• Overlappingà 1,2
Genes after DFPà 635withSkip Factor 1,Zeta0.35,piVal 0.65andOverlapping 1
![Page 9: Pathway based OMICs data classification](https://reader031.fdocuments.us/reader031/viewer/2022030317/5a6532ee7f8b9a76138b475d/html5/thumbnails/9.jpg)
Enrichment
• BreastCancer• Numberofpathwaysselectedthroughenrichment:1585• Numberofpathwayswithmorethan10genes:859
• Glioma• Numberofpathwaysbyenrichment:1612p-value0.0001• First1000pathwayswithmorethan10genesandlowestp-value
![Page 10: Pathway based OMICs data classification](https://reader031.fdocuments.us/reader031/viewer/2022030317/5a6532ee7f8b9a76138b475d/html5/thumbnails/10.jpg)
ClassificationwithSVM
• Linear SVM
• Two level cross-validation• 3outer folds• 2inner folds
• C:1e-5,1e-4,1e-3,1e-2,1e-1,1e0,1e1,1e2,1e3,1e4,1e5,1e6
![Page 11: Pathway based OMICs data classification](https://reader031.fdocuments.us/reader031/viewer/2022030317/5a6532ee7f8b9a76138b475d/html5/thumbnails/11.jpg)
FirststepofclassificationPatients
gene
spathway
1
gene
spathway
2
gene
spathway
i
gene
spathway
3
LinearSVM1
LinearSVM2
LinearSVM3
LinearSVMi
Classprob.
Patie
nts
Patie
nts
Patie
nts
Patie
nts
Classprob.
Classprob.
Classprob.
![Page 12: Pathway based OMICs data classification](https://reader031.fdocuments.us/reader031/viewer/2022030317/5a6532ee7f8b9a76138b475d/html5/thumbnails/12.jpg)
Sizevs PathwaysAccuracy(BC)
Correlation:0.028
![Page 13: Pathway based OMICs data classification](https://reader031.fdocuments.us/reader031/viewer/2022030317/5a6532ee7f8b9a76138b475d/html5/thumbnails/13.jpg)
SizevsPathwaysAccuracy(Glioma)
Correlation:0.342
![Page 14: Pathway based OMICs data classification](https://reader031.fdocuments.us/reader031/viewer/2022030317/5a6532ee7f8b9a76138b475d/html5/thumbnails/14.jpg)
Pathwaysafterpermutationtest
• 1000permutationtestsonthepathwayswithbestaccuracies
• Breastcancer• Numberofpathwaysthatpassedpermutationtest:36
• Lowestaccuracy77.9%• Highestaccuracy84.6%
• Glioma• Numberofpathwaysthatpassedpermutationtest:278
• Lowestaccuracy80%• Highestaccuracy88%
![Page 15: Pathway based OMICs data classification](https://reader031.fdocuments.us/reader031/viewer/2022030317/5a6532ee7f8b9a76138b475d/html5/thumbnails/15.jpg)
• ACEVEDO_FGFR1_TARGETS_IN_PROSTATE_CANCER_MODEL_UP
• DEBIASI_APOPTOSIS_BY_REOVIRUS_INFECTION_DN
• DELACROIX_RARG_BOUND_MEF
• ENK_UV_RESPONSE_EPIDERMIS_UP
• ENK_UV_RESPONSE_KERATINOCYTE_DN
• FARMER_BREAST_CANCER_APOCRINE_VS_BASAL
• GO_CELLULAR_RESPONSE_TO_LIPID
• GO_CIRCULATORY_SYSTEM_PROCESS
• GO_GLAND_DEVELOPMENT
• GO_REGIONALIZATION
• GO_REGULATION_OF_CELL_CYCLE_PHASE_TRANSITION
• GO_REGULATION_OF_PROTEIN_SERINE_THREONINE_KINASE_ACTIVITY
• GO_RESPONSE_TO_ALCOHOL
• GO_RESPONSE_TO_ESTROGEN
• GO_RESPONSE_TO_STEROID_HORMONE
• GO_UROGENITAL_SYSTEM_DEVELOPMENT
• GSE1460_NAIVE_CD4_TCELL_ADULT_BLOOD_VS_THYMIC_STROMAL_CELL_DN
• GSE21927_SPLEEN_VS_4T1_TUMOR_MONOCYTE_BALBC_DN
• GSE23502_WT_VS_HDC_KO_MYELOID_DERIVED_SUPPRESSOR_CELL_COLON_TUMOR_DN
• GSE26351_WNT_VS_BMP_PATHWAY_STIM_HEMATOPOIETIC_PROGENITORS_UP
• HALLMARK_ESTROGEN_RESPONSE_LATE
• LEI_MYB_TARGETS
• LIU_PROSTATE_CANCER_DN
• MODULE_18
• MODULE_255
• MODULE_52
• NFE2L2.V2
• SATO_SILENCED_BY_METHYLATION_IN_PANCREATIC_CANCER_1
• SHEN_SMARCA2_TARGETS_DN
• SMID_BREAST_CANCER_RELAPSE_IN_BONE_DN
• V$ALPHACP1_01
• V$TEF1_Q6
• V$ZIC2_01
• VANTVEER_BREAST_CANCER_ESR1_DN
• VECCHI_GASTRIC_CANCER_EARLY_DN
• ZHANG_BREAST_CANCER_PROGENITORS_UP
BCPathways
![Page 16: Pathway based OMICs data classification](https://reader031.fdocuments.us/reader031/viewer/2022030317/5a6532ee7f8b9a76138b475d/html5/thumbnails/16.jpg)
GliomaPathways
• MEISSNER_NPC_HCP_WITH_H3K4ME2 • YYCATTCAWW_UNKNOWN • RIGGI_EWING_SARCOMA_PROGENITOR_UP • DEURIG_T_CELL_PROLYMPHOCYTIC_LEUKEMIA_DN
• MODULE_169 • GO_REGULATION_OF_MEMBRANE_POTENTIAL • GSE24574_BCL6_LOW_TFH_VS_NAIVE_CD4_TCELL_UP
• GSE25677_MPL_VS_R848_STIM_BCELL_DN • REACTOME_AXON_GUIDANCE • MODULE_19• HELLER_HDAC_TARGETS_SILENCED_BY_METHYLATION_UP
• GO_ACTIN_BINDING• GSE3982_EOSINOPHIL_VS_BASOPHIL_UP• GSE3982_MAC_VS_TH2_UP • V$TATA_C • GO_REGULATION_OF_ANATOMICAL_STRUCTURE_SIZE
• MODULE_52 • SCHAEFFER_PROSTATE_DEVELOPMENT_48HR_UP•DAVICIONI_TARGETS_OF_PAX_FOXO1_FUSIONS_UP
•DEURIG_T_CELL_PROLYMPHOCYTIC_LEUKEMIA_UP
• GSE21063_CTRL_VS_ANTI_IGM_STIM_BCELL_NFATC1_KO_16H_UP
• KAECH_NAIVE_VS_MEMORY_CD8_TCELL_DN• SANSOM_APC_TARGETS_DN • GO_SINGLE_ORGANISM_CELL_ADHESION• HOLLMANN_APOPTOSIS_VIA_CD40_DN
![Page 17: Pathway based OMICs data classification](https://reader031.fdocuments.us/reader031/viewer/2022030317/5a6532ee7f8b9a76138b475d/html5/thumbnails/17.jpg)
• GSE22025_TGFB1_VS_TGFB1_AND_PROGESTERONE_TREATED_CD4_TCELL_DN
• GO_CELL_SUBSTRATE_JUNCTION
• GSE3982_NEUTROPHIL_VS_EFF_MEMORY_CD4_TCELL_UP
• HIRSCH_CELLULAR_TRANSFORMATION_SIGNATURE_UP
• GSE21927_SPLENIC_C26GM_TUMOROUS_VS_BONE_MARROW_MONOCYTES_DN
• GSE3982_BASOPHIL_VS_CENT_MEMORY_CD4_TCELL_UP
• MCBRYAN_PUBERTAL_BREAST_4_5WK_UP
• GO_CELL_CELL_JUNCTION
• GSE13411_NAIVE_BCELL_VS_PLASMA_CELL_UP
• GO_AXON
• GO_REGULATION_OF_INTRACELLULAR_PROTEIN_TRANSPORT
• GO_TELENCEPHALON_DEVELOPMENT
• GSE13484_UNSTIM_VS_12H_YF17D_VACCINE_STIM_PBMC_DN
• LEF1_UP.V1_DN
• CASORELLI_ACUTE_PROMYELOCYTIC_LEUKEMIA_UP
• GO_ACTIVATION_OF_IMMUNE_RESPONSE
• GO_EPITHELIAL_CELL_DIFFERENTIATION
• GO_POSITIVE_REGULATION_OF_CELL_ADHESION
• GSE15735_2H_VS_12H_HDAC_INHIBITOR_TREATED_CD4_TCELL_UP
• MODULE_8
• BLALOCK_ALZHEIMERS_DISEASE_INCIPIENT_UP
• GO_DENDRITE
• GSE3982_CENT_MEMORY_CD4_TCELL_VS_TH2_UP
• KIM_WT1_TARGETS_UP
• GO_REGULATION_OF_NEURON_PROJECTION_DEVELOPMENT
GliomaPathways
![Page 18: Pathway based OMICs data classification](https://reader031.fdocuments.us/reader031/viewer/2022030317/5a6532ee7f8b9a76138b475d/html5/thumbnails/18.jpg)
Graph- (1)
• BuildinteractiongraphbetweenpathwayandmiRNAcommunities
• Wefirstcomputeinteractionsbetweenpathways• InteractionScorematrix
• WethenaddmiRNAsconnectingthemtopathways• CorrelationmatrixbetweenmiRNAsandgenes• Fisher'sexacttest
• WeaddedgesbetweenmiRNAs• Weightednetworkprojection
![Page 19: Pathway based OMICs data classification](https://reader031.fdocuments.us/reader031/viewer/2022030317/5a6532ee7f8b9a76138b475d/html5/thumbnails/19.jpg)
Graph- (2)
• GroupmiRNAsincommunities• Walktrap algorithm
• WereplacemiRNAswithnodesrepresentingmiRNAcommunities
• Wefinallyidentifycommunitiesinthewholeinteractiongraph
![Page 20: Pathway based OMICs data classification](https://reader031.fdocuments.us/reader031/viewer/2022030317/5a6532ee7f8b9a76138b475d/html5/thumbnails/20.jpg)
InteractionScore
Relationsamongpathways:interactionscore(IS)
!" = |%& −%(|"& + "(
MandSarerespectivelymeanandstandarddeviationofthetwopathwaysxandy
Weapplyacutoffontheresultinginteractionmatrix
![Page 21: Pathway based OMICs data classification](https://reader031.fdocuments.us/reader031/viewer/2022030317/5a6532ee7f8b9a76138b475d/html5/thumbnails/21.jpg)
miRNAandPathwaysinteraction
• WeevaluatePearsoncorrelationbetweenthemiRNAandallthegenesinthepathway.Wethenapplyacutofftoselectstrongcorrelations.
• ThenforeachmiRNAandpathwayweuseFisher’sexacttest,todetermineifthemiRNAissignificantlylinkedtothepathway(i.e.wecheckifthereisasignificantnumberofgenesincommon)
![Page 22: Pathway based OMICs data classification](https://reader031.fdocuments.us/reader031/viewer/2022030317/5a6532ee7f8b9a76138b475d/html5/thumbnails/22.jpg)
TheGoal
![Page 23: Pathway based OMICs data classification](https://reader031.fdocuments.us/reader031/viewer/2022030317/5a6532ee7f8b9a76138b475d/html5/thumbnails/23.jpg)
TheGoal(2)
![Page 24: Pathway based OMICs data classification](https://reader031.fdocuments.us/reader031/viewer/2022030317/5a6532ee7f8b9a76138b475d/html5/thumbnails/24.jpg)
FinalclassificationPatients
Stacking
with
pathw
ay
SVM Classes foreach patient
Patients
gene
spathways
LinearSVM
LinearSVM
LinearSVM
LinearSVM
Classprob.
Patie
nts
Patients
miRNA
sand
pathw
ayconn
ected
LinearSVM
LinearSVM
LinearSVM
LinearSVM
Classprob.
Patie
nts
Patients
Stacking
with
pathway
andmiRNA
s
SVM Classes foreach patient
![Page 25: Pathway based OMICs data classification](https://reader031.fdocuments.us/reader031/viewer/2022030317/5a6532ee7f8b9a76138b475d/html5/thumbnails/25.jpg)
Fine