Pathway analysis computing system and method

Patent metadata

Field	Value
Publication number	US-11043282-B2
Application number	US-201715667535-A
Country	US
Kind code	B2
Filing date	Aug 2, 2017
Priority date	Apr 29, 2010
Publication date	Jun 22, 2021
Grant date	Jun 22, 2021

Abstract

The present invention relates to methods for evaluating the probability that a patient's diagnosis may be treated with a particular clinical regimen or therapy.

First claim

What is claimed is:

1. A pathway analysis computing system comprising:
a computer readable, non-transitory memory storing software instructions and a probabilistic graphical model (PGM) of a gene expression pathway network, the PGM including:
directed graphs that include pathway interactions and that include hidden cell states and observed cell states of biological cellular entities for a set of genes, wherein the biological cellular entities include gene copy number, mRNA expression, protein level, and protein activity for each of the set of genes, and wherein measurements are available for the observed cell states and not available for the hidden cell states;
wherein the directed graphs form a data structure including nodes representing the biological cellular entities and edges representing the pathway interactions among the nodes of the biological cellular entities associated with the pathway interactions; and
wherein variables of the directed graphs represent differential states of the hidden cell states and the observed cell states relative to corresponding normal levels of the biological cellular entities, and where factors of edges of the directed graphs are assigned to the edges representing the pathway interactions between the biological cellular entities, and wherein the pathway interactions are biochemical interactions; and

a processor coupled with the memory and configured, upon execution of the software instructions, to output integrated pathway activities for pathways associated with a patient by:
initializing the PGM by assigning, in the memory, values measured from a tissue sample of the patient to the observed cell states of observed nodes in the directed graphs;
simulating the gene expression pathway network, using the initialized PGM, to estimate the hidden cell states of hidden nodes of the directed graphs for the tissue sample based on the assigned observed cell states in the directed graphs and according to the factors of edges of the directed graphs that connect the hidden nodes and the observed nodes, wherein the simulating includes:
determining a probability of a given hidden cell state for each hidden node of the hidden nodes based on (1) probabilities of cell states of parent nodes of the hidden node and (2) the factors of edges from the parent nodes, and
iteratively updating the hidden cell states based on the probabilities of the cell states and the factors of edges; and

outputting, from the memory, the estimated hidden cell states of the biological cellular entities for the tissue sample as integrated pathway activities of the tissue sample, wherein the estimated hidden cell states include a protein activity.

2. The system of claim 1 , wherein the probabilistic graphical model comprises at least two directed graphs.

3. The system of claim 1 , wherein the directed graphs comprise a factor graphs.

4. The system of claim 1 , wherein the directed graphs comprise a Bayesian network.

5. The system of claim 1 , wherein the pathways are formatted according to BioPAX formatting rules.

6. The system of claim 5 , wherein the pathways are formatted according to BioPAX level 2.

7. The system of claim 1 , wherein the biological cellular entities include at least four different biological cellular entities.

8. The system of claim 1 , wherein the integrated pathway activities include at least three states for the biological cellular entities.

9. The system of claim 8 , wherein the at least three states include: an active state, an inactive state, and a nominal state.

10. The system of claim 1 , wherein the estimated hidden cell states are estimated by belief propagation.

11. The system of claim 1 , wherein the estimated hidden cell states are estimated by iterating between inferring probabilities of the hidden cell states and updating the hidden cell states to maximize likelihood given probabilities of the hidden cell states.

12. The system of claim 11 , wherein the estimated hidden cell states are estimated via an implementation of an expectation-maximization algorithm.

13. The system of claim 1 , wherein the factors between the biological cellular entities include non-negative functions.

14. The system of claim 13 , wherein the non-negative functions constrain values of the cell states.

15. The system of claim 1 , wherein the factors define probabilities associated with states of the biological cellular entities.

16. The system of claim 1 , wherein the gene expression pathway network includes at least one of: a regulatory pathway network and a signaling pathway network.

17. The system of claim 1 , wherein the values obtained from the tissue sample of the patient and assigned to the observed cell states include at least one of the following: mutation, a differential genetic sequence object, a gene copy number, a transcription level, a translation level, a protein activity, and a protein interaction.

18. The system of claim 1 , wherein the tissue sample is at least one of the following: a diseased tissue, a normal tissue, an aging tissue, a recovering tissue, an animal tissue, an experimentally treated tissue, a cancer tissue, and a patient fluid.

19. The system of claim 1 , wherein the PGM is generated using observed cell states of a cohort of patients that have a same condition as the patient.

20. A method comprising:
measuring values from a tissue sample of a patient, wherein the measured values correspond to observed cell states and not to hidden cell states;
storing, by a computer in a database, a probabilistic graphical model (PGM) of a gene expression pathway network, the PGM including:
directed graphs that include pathway interactions and that include hidden cell states and observed cell states of biological cellular entities for a set of genes, wherein the biological cellular entities include gene copy number, mRNA expression, protein level, and protein activity for each of the set of genes, and wherein measurements are available for the observed cell states and not available for the hidden cell states;
wherein the directed graphs form a data structure including nodes representing the biological cellular entities and edges representing the pathway interactions among the nodes of the biological cellular entities associated with the pathway interactions; and
wherein variables of the directed graphs represent differential states of the hidden cell states and the observed cell states relative to corresponding normal levels of the biological cellular entities, and where factors of edges of the directed graphs are assigned to the edges representing the pathway interactions between the biological cellular entities, and wherein the pathway interactions are biochemical interactions;

initializing, by the computer, the PGM by assigning values measured from the tissue sample of the patient to the observed cell states of observed nodes in the directed graphs;
simulating, by the computer, the gene expression pathway network, using the initialized PGM, to estimate the hidden cell states of hidden nodes of the directed graphs for the tissue sample based on the assigned observed cell states in the directed graphs and according to the factors of edges of the directed graphs that connect the hidden nodes and the observed nodes, wherein the simulating includes:
determining a probability of a given hidden cell state for each hidden node of the hidden nodes based on (1) probabilities of cell states of parent nodes of the hidden node and (2) the factors of edges from the parent nodes, and
iteratively up

Assignees

• Univ California
• The Regents Of The Unviersity Of California

Inventors

• Vaske Charles J
• Benz Stephen C
• Stuart Joshua M
• Haussler David

Classifications

• G16B40/30

  Unsupervised data analysis · CPC title

• G16B5/30 Primary

  Dynamic-time models · CPC title

• G16B25/10

  Gene or protein expression profiling; Expression-ratio estimation or normalisation · CPC title

• G16B40/00

  ICT specially adapted for biostatistics; ICT specially adapted for bioinformatics-related machine learning or data mining, e.g. knowledge discovery or pattern finding · CPC title

• G16B25/00

  ICT specially adapted for hybridisation; ICT specially adapted for gene or protein expression · CPC title