Bayesian Decomposition Expression to Pathways Michael Ochs - - PowerPoint PPT Presentation

▶
bayesian decomposition expression to pathways
SMART_READER_LITE
LIVE PREVIEW

Bayesian Decomposition Expression to Pathways Michael Ochs - - PowerPoint PPT Presentation

May 01, 2023 282 likes •616 views

Bayesian Decomposition Expression to Pathways Michael Ochs Bioinformatics Group Fox Chase Cancer Center Cancer Biology Cancer is many Diseases but with a Single Theme a cell becomes immortal Insert poster a cell becomes mobile

slide-1
SLIDE 1

Bioinformatics Group Fox Chase Cancer Center

Bayesian Decomposition Expression to Pathways

Michael Ochs

slide-2
SLIDE 2

Bioinformatics Group Fox Chase Cancer Center

Cancer Biology

Insert poster Cancer is many Diseases but with a Single Theme

  • a cell becomes immortal
  • a cell becomes mobile

Signalling and Metabolic Pathways Hold the Key

www.biosource.com

slide-3
SLIDE 3

Bioinformatics Group Fox Chase Cancer Center

Signalling Pathways

Downward, Nature, 411, 759, 2001

mRNA Stimulus Signal Transduction Transcription

slide-4
SLIDE 4

Bioinformatics Group Fox Chase Cancer Center

Identifying Pathways

mRNA Interacting Pathways Lead to Confusion if All Genes Need to Lie in a Single Cluster

www.promega.com

slide-5
SLIDE 5

Bioinformatics Group Fox Chase Cancer Center

Bayesian Decomposition

  • Data Mining/Pattern Recognition Algorithm

– Unsupervised Method – Create Multiple, Overlapping “Clusters”

  • Each Gene can be in Multiple Patterns
  • Get to Pathways: Key for Cancer Development
  • Methodology

– Markov Chain Monte Carlo Algorithm – Simulated Annealing – Integration of Prior Knowledge

slide-6
SLIDE 6

Bioinformatics Group Fox Chase Cancer Center

BD: Matrix Decomposition

* * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * *

Data X

gene 1 gene N * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * gene 1 gene N pattern 1 pattern k Exp 1 Exp M * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * pattern 1 pattern k

Distribution of Patterns Patterns of Behavior =

The behavior of

  • ne gene can be

explained as a mixture of patterns with different behaviors Exp 1 Exp M

slide-7
SLIDE 7

Bioinformatics Group Fox Chase Cancer Center

BD: Domains

A Atomic Domain P Atomic Domain * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * *

A P Model Domain

* * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * *

Data X Data Domain

convolution convolution

slide-8
SLIDE 8

Bioinformatics Group Fox Chase Cancer Center

BD: Markov Chain MC

Cloud in N-Dimensional Space: Probability Density for the Model Results from Atomic Domain Prior, Model Functions (Prior), and the Likelihood

Based on Maximum Entropy Data Consultants Massive Inference Sampler

slide-9
SLIDE 9

Bioinformatics Group Fox Chase Cancer Center

BD Requirements

  • Data Points > (A + P) Points
  • Atomic Domains (Sibisi and Skilling, J R Stat Soc B, 59, 217, 1997)

– Positive Additive Distributions – Infinitely Divisible Process

  • Model Domains

– Linked to Atomic Domains by Model Function – Correlations between Parameters are Introduced by Model Functions (Atomic > Model)

slide-10
SLIDE 10

Bioinformatics Group Fox Chase Cancer Center

BD Features

  • Basis Vectors (Patterns) are Nonorthogonal

– Physically Meaningful if Good Model – Artifacts Removed if Do Not Fit Model

  • Noise is Treated

– Noise is Integral Part of Fitting Process – Artifacts Often Appear in Residuals (i.e. noise)

  • Markov Chain Sampling Yields

– Mean of Probable Distributions and Patterns – Uncertainties for Distributions and Patterns

slide-11
SLIDE 11

Bioinformatics Group Fox Chase Cancer Center

BD: Gene Expression

A Atomic Domain P Atomic Domain

* * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * gene 1 gene N pattern 1 pattern k

*

* * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * pattern 1 pattern k Exp 1 Exp M

*

slide-12
SLIDE 12

Bioinformatics Group Fox Chase Cancer Center

Rosetta Data Set

  • Filtering

– Eliminate Genes

  • >25% Data Missing in Ratios or Uncertainties
  • < 2 Experiments with 3 Fold Change

– Eliminate Experiments

  • < 2 Genes Changing by 3 Fold
  • Uncertainties

– Used Values from Rosetta Error Model – Missing Data Log Ratio =1, Log Unc = 100

slide-13
SLIDE 13

Bioinformatics Group Fox Chase Cancer Center

Analysis

  • Analyzed Full Experimental Data with PCA

– Estimate of Dimensionality of Data

  • Bayesian Decomposition

– Filtered Data: 764 Genes, 228 Experiments – Ran Multiple Seeds, Multiple Pattern Number – Focus on Dimensions Suggested by PCA

  • Data Driven

– Let Analysis Determine Where to Look

slide-14
SLIDE 14

Bioinformatics Group Fox Chase Cancer Center

PCA Results

Principal Component Score (EigenValue)

slide-15
SLIDE 15

Bioinformatics Group Fox Chase Cancer Center

Bayesian Decomposition

  • Distributions

– Assignment of Genes to Patterns

  • Patterns

– Each Pattern Defines Behavior Across Experiments

  • Experimental Patterns

– Experiments explained by a single pattern – Correlations between experiments

  • Genes in Patterns

– Identify biological processes – Identify correlations in genes

slide-16
SLIDE 16

Bioinformatics Group Fox Chase Cancer Center

Experiments High in One Pattern

  • Pattern 1

– YHR034C 56%

  • Pattern 2

– rpd3 89%

  • Pattern 3

– ssn6 (cyc8) 76% – YER024W 56% – tup1 54%

  • Pattern 5

– YJL107C 53% – yap3 51%

slide-17
SLIDE 17

Bioinformatics Group Fox Chase Cancer Center

Genes in Patterns

(Proteome Database Cellular Role)

  • Pattern 1

– 403 Genes – 22/36 AA metabolism – 9 additional metabolism

  • Pattern 2

– 410 Genes – 7/27 metabolism – 7/27 DNA/RNA processing – 6 transport

  • Pattern 3

– 390 Genes – 13/26 metabolism – 6 transport, 4 Pol II

  • Pattern 4

– 276 Genes, 30/50 Unknown

  • Pattern 5

– 355 Genes – 14/37 carbohydrate metabolism – 7/37 cell stress – 6 transport

  • Pattern 6

– 297 Genes, 30/50 unknown

  • Pattern 7

– 223 Genes – 13/23 mating response – 5/23 meiosis

AA Pattern Carbo Pattern Metabolic Pattern Mating Pattern

slide-18
SLIDE 18

Bioinformatics Group Fox Chase Cancer Center

Metabolic Patterns

  • Patterns 1 and 5

– yap 3 98% – YJL107C 98% – YHR034C 98% – FR901,228 98%

  • Patterns 1, 3, and 5

– ssn6 100% – swi6 99% – yap 3 98% – YJL107C 98% – YHR034C 98% – FR901,228 98%

slide-19
SLIDE 19

Bioinformatics Group Fox Chase Cancer Center

Metabolic Pattern

0% 10% 20% 30% 40% 50% 60% 70% 80% ssn6 (haploid) tup1 (haploid) yer024w AA Metab Carbo Mating

Behavior Explained by Pattern

slide-20
SLIDE 20

Bioinformatics Group Fox Chase Cancer Center

Sterile Family Proteins

0% 5% 10% 15% 20% 25% 30% 35% 40% 45%

AA Patt 2 Metab Patt 4 Carbo Patt 6 Mating

ste11 (haploid) ste12 (haploid) ste18 (haploid) ste2 (haploid) ste20 (**11) ste24 (haploid) ste4 (haploid) ste5 (haploid) ste7 (haploid)

Behavior Explained by Pattern

slide-21
SLIDE 21

Bioinformatics Group Fox Chase Cancer Center

Ste2

0% 5% 10% 15% 20% 25% 30% 35% 40% 45%

AA Patt 2 Metab Patt 4 Carbo Patt 6 Mating

ste2 (haploid) yil117c (haploid)

YIL117C is prm5, a pheromone regulated protein of unknown function

Behavior Explained by Pattern

slide-22
SLIDE 22

Bioinformatics Group Fox Chase Cancer Center

Mating Pattern

0% 5% 10% 15% 20% 25% 30% 35% 40% dig1, dig2 (haploid) dig1, dig2 dig1 dig2 ste20 (**11) fus3 (haploid)

AA Metab Carbo Mating

Behavior Explained by Pattern

slide-23
SLIDE 23

Bioinformatics Group Fox Chase Cancer Center

Mating Pathway

Posas, et al, Curr Opin Microbiology, 1, 175, 1998

slide-24
SLIDE 24

Bioinformatics Group Fox Chase Cancer Center

Mating Pattern

0% 5% 10% 15% 20% 25% 30% 35% 40% 45% 50% fus3 (haploid) fus3, kss1 (haploid) ste11 (haploid) ste7 (haploid) ste5 (haploid) ste12 (haploid)

AA Metab Carbo Mating 6

Behavior Explained by Pattern

slide-25
SLIDE 25

Bioinformatics Group Fox Chase Cancer Center

Correlations (Fus3/Kss1)

0.5 1 1.5 2 2.5 3

YAL012W YAL034W YAL066W YAR009C YAR047C YAR070C YBL005W YBL043W YBL049W YBL098W YBL101W YBR012C YBR012W YBR040W

fus3 (haploid) fus3, kss1 (haploid) Ratio of Expression

slide-26
SLIDE 26

Bioinformatics Group Fox Chase Cancer Center

Mating Pattern Correlations

0.5 1 1.5 2 2.5 3 YAL018C YMR082C MEI5 YFR057W STE2 ** PES4 FIG1 ** YAL066W SPO19 YNL018C YNL028W YOR235W AGA2 ** YMR082C FUS1 YOR376W YMR082C SST2 ** BAR1 ** YMR082C YOL131W PRR2 AGA1 ** YER181C YPL280W YLR042C HOP2 TEC1

fus3 (haploid) fus3, kss1 (haploid)

* * * * * *

Ratio of Expression

* * * * *

slide-27
SLIDE 27

Bioinformatics Group Fox Chase Cancer Center

Pattern 6

  • Mating Pattern -> Pattern 6 for Ste11, Ste7,

Ste5, and Ste12

  • PSI-BLAST and SMART pick up 5

matches among unknown ORFs to transposon and retroposon proteins

slide-28
SLIDE 28

Bioinformatics Group Fox Chase Cancer Center

Conclusions

  • Life is Very Complex

– Multiple Pathways and Interactions for Each Protein with Transcription/Translation – Natural Stochastic Variations

  • Analysis Tools Must

– Isolate Areas of Interest without Loss of Knowledge Discovery – Incorporate Maximal Prior Knowledge to Reduce “Search Space”

slide-29
SLIDE 29

Bioinformatics Group Fox Chase Cancer Center

Bayesian Decomposition

  • Ability to Identify Overlapping

Coexpression Can Lead to Identification of Pathways Affected in an Experiment

  • Capability to Encode Prior Knowledge Can

Improve Results

– Ochs et al, J Magn Res, 137, 161, 1999 – Ochs et al, Magn Res Med, in press

slide-30
SLIDE 30

Bioinformatics Group Fox Chase Cancer Center

Future Directions

  • Application to Clinical Trials

– Study of GIST Tumor Response to Gleevec – Study of Drug Response of HOSE Lines

  • Bayesian Decomposition Development

– Incorporation of Time Domain Modelling – Incorporation of Knowledge of Coexpression

slide-31
SLIDE 31

Bioinformatics Group Fox Chase Cancer Center

Credits (Definitely Due)

  • Fox Chase

– Frank Manion – Thomas Moloshok – Jeffrey Grant – Yue Zhang – Ghislain Bidaut – Burt Eisenberg – Andy Godwin

  • City of Hope

– Bob Klevecz

  • Johns Hopkins

– Giovanni Parmigiani

  • Fox Chase (NMR)

– Truman Brown (Columbia)