Enabling Patient-Driven Medicine Using Graph Database
Download as PPTX, PDF0 likes102 views
The document discusses using a graph database approach to integrate diverse types of patient omics and clinical data in order to enable a systems biology perspective in medicine. It describes building a graph database called GENOME to integrate multi-omics and clinical data from brain tumor patients to identify novel biomarkers and survival correlates. Random walk simulations on the graph are able to find statistically significant pathways involving different data types that may be biologically meaningful. This approach allows an unbiased and comprehensive analysis not possible with traditional methods.
Enabling Patient-Driven Medicine Using Graph Database
- 1. © 2022 Neo4j, Inc. All rights reserved. Enabling Patient-Driven Medicine Using Graph Database A systems biology paradigm for medicine Kasthuri Kannan, PhD Associate Professor, Director (Computational Pathology Program) Departments of Translational Molecular Pathology & Neurosurgery
- 2. © 2022 Neo4j, Inc. All rights reserved. Internship Knowledge Based Systems Jan. 2007-Dec. 2007 Research Specialist/Associate Stowers Institute for Medical Research/ Penn State Jan. 2008--Mar. 2011 Research Fellow/Associate Memorial Sloan-Kettering Cancer Center Apr. 2011-Oct. 2013 Assistant Professor New York University Nov. 2013-Aug. 2019 Bachelor of Science Mathematics (BSc) University of Madras Apr. 1998 Master of Science Mathematics (MSc) IIT, Madras Apr. 2000 Master of Science Mathematics (MS) Texas A&M University Apr. 2002 education acquired credentials Kasthuri Kannan Doctor of Philosophy Computer Science (PhD) Texas A&M University Aug. 2008 creativity skills Several successful collaborations Active learner and listener Emotional intelligence Communication Proven track record of applying data science to biology and omics data Image analysis/processing Designing/directing data science & machine learning courses Software development Bioinformatics pipeline development NGS analysis/processing Systems Biology Directed insights using data science Mathematical & statistical modeling + Visualization Data science (14 years) Graph Databases (Neo4j) Machine learning Several bioinformatics tools Implementation skills R/Python/C/C++ Strong Unix skills Image analysis Statistics, Mathematics Top publications work experience directed efforts Associate Professor & Director UT MD Anderson Cancer Center Feb. 2020-Present self- motivation Data Scientist (Cancer Research)
- 3. © 2022 Neo4j, Inc. All rights reserved. 3 The state of modern biomedicine
- 4. © 2022 Neo4j, Inc. All rights reserved. 4 Basics (Genetics) Gene Expression. Image Credit: udaix/Shutterstock.com Gene expression
- 5. © 2022 Neo4j, Inc. All rights reserved. 5 Same DNA but different mRNA (and hence different proteins) Image Credits: Hubrecht Institute, https://www.pnas.org/doi/10.1073/pnas.0509380102
- 6. © 2022 Neo4j, Inc. All rights reserved. 6 How different mRNAs are produced by the same DNA? A UGA Protein X Protein Y Image Credits: courses.lumenlearning.com, https://www.clancymedicalgroup.com/amino-acids/, themedicalbiochemistrypage.org
- 7. © 2022 Neo4j, Inc. All rights reserved. 7
- 8. © 2022 Neo4j, Inc. All rights reserved. 8 Yet another mechanism… Genes Expressed Image Credit: https://www.zymoresearch.com/pages/what-is-epigenetics
- 9. © 2022 Neo4j, Inc. All rights reserved. 9 Nature vs Nurture Image Credit: https://www.genomicseducation.hee.nhs.uk/
- 10. © 2022 Neo4j, Inc. All rights reserved. 10 Absence of consensus
- 11. © 2022 Neo4j, Inc. All rights reserved. 11 Algorithm • Find all 3.2 billion letters as a sequence • or extract sequential coding letters of the genome ~ 20K genes • Look for gene sequence changes • Look for methylated regions (promotors) • Correlate with the mRNA numbers (transcripts)
- 12. © 2022 Neo4j, Inc. All rights reserved. 12 Next-generation sequencing Made identifying the sequences of nucleotides possible Can identify both DNA, RNA and methylated nucleotides/sequences Affordable with less turn-around time (< $1000 & 8-11 days) Reference genome available through Human Genome Project Image Credit: https://bitesizebio.com/
- 13. © 2022 Neo4j, Inc. All rights reserved. 13 Data deluge in biology • Explosion of biological datasets Image Credits: https://www.nature.com/articles/d41586-019-03536-x, https://th.wikipedia.org/wiki/%E0%B9%84%E0%B8%9F%E0%B8%A5%E0%B9%8C:Karyotype_color_chromosomes_white_b ackground.png, https://www.activemotif.com/catalog/2/gene-regulation, https://www.activemotif.com/catalog/42/dna- methylation-enrichment
- 14. © 2022 Neo4j, Inc. All rights reserved. 14 Yet no consensus Image Credit: https://www.flyaeroguard.com/learning-center/parts-of-an-airplane/
- 15. © 2022 Neo4j, Inc. All rights reserved. 15 Systems thinking vs reductionism in biology • Modern biology is more or less “reductionistic” • Divide and conquer may not work when studying complex systems • Biomedicine and especially cancer is complex and adaptive • Reductionism is like relational databases (breaking into tables) • Need an integrative paradigm – SYSTEMS APPROACH
- 16. © 2022 Neo4j, Inc. All rights reserved. 16 Integrativetechnology defines systems biology Image Credit: Institute of Systems Biology
- 17. © 2022 Neo4j, Inc. All rights reserved. 17 Systems biology initiative for brain tumors Goal of the project: creating a unifying computational framework to integrate all available patient-centric brain tumor recurrence data to generate hypotheses in an unbiased manner for actionable target validation and predicting disease outcomes for initiating clinical trials.
- 18. © 2022 Neo4j, Inc. All rights reserved. 18 Glioblastoma – most lethal of several brain tumors Image Credit: braintumor.org
- 19. © 2022 Neo4j, Inc. All rights reserved. 19 Graph architecture • Integration and unbiased analysis becomes natural • Integration is handled by defining the relationships and unbiased analysis is handled by writing appropriate queries
- 20. © 2022 Neo4j, Inc. All rights reserved. 20 Is this idea new? 1Irina Balaur, Mansoor Saqi, Ana Barat, Artem Lysenko, Alexander Mazein, Christopher J. Rawlings, Heather J. Ruskin, and Charles Auffray. EpiGeNet: A Graph Database of Interdependencies Between Genetic and Epigenetic Events in Colorectal Cancer. Journal of Computational Biology. Oct 2017.969-980 Integration is either subjective, restricted to datatypes or phenotypes, i.e., relationships are not general enough Necessity to create additional relationship types for new data Results in restricted survival analysis
- 21. © 2022 Neo4j, Inc. All rights reserved. 21 What is new? Molecular Event patients > 1 Our database (Basic Representation) Molecular Event
- 22. © 2022 Neo4j, Inc. All rights reserved. 22 Graph ENgine fOr systeMs mEdicine (GENOME) Such architecture allows any type of patient-data integration and allows comprehensive survival analysis Copy number events Expression events Mutation events Methylation events Any two events are connected if they are shared by more than one patient Graph ENgine fOr systeMs mEdicine (GENOME)
- 23. © 2022 Neo4j, Inc. All rights reserved. 23 GENOME overview
- 24. © 2022 Neo4j, Inc. All rights reserved. 24 Filters fastq files align .bam files re-align (GATK) variant calling filtering driver analysis driver mutations Molecular Event patients > 1 Molecular Event
- 25. © 2022 Neo4j, Inc. All rights reserved. 25 Survival correlation A B C 13 w x y z (wx), (xy), (yz), (zw): 1-length paths (4) (wxy), (xyz), (yzw), (zwx): 2-length paths (4) (wxyz), (xyzw), (yzwx), (zwxy): 3-length paths (4) (wxyzw): 4-length paths (1) 11K nodes and 16M relationships
- 26. © 2022 Neo4j, Inc. All rights reserved. 26 Random walk simulations I Path no. 51 p < 0.03 ITGA8-ABCA4-C22orf42- Intergenic CpG-NGF- CCL1-HBBP1-14q32.13- 10q11.22-C17orf54 Path no. 54 Path no. 51 CAMTA1-17q25.3-10q26.13- WWOX-Intergenic CpG- CHST1-17q25o3-Intergenic CpG-TRRAP-CPS1 Path no. 54 p < 0.03
- 27. © 2022 Neo4j, Inc. All rights reserved. 27 Random walk simulations II p < 0.005
- 28. © 2022 Neo4j, Inc. All rights reserved. 28 Random walk simulations III NMUR2-GABRG2-FKBP8- KCNK9-HOXD1-6p22.3 p < 0.001 Path no. 16
- 29. © 2022 Neo4j, Inc. All rights reserved. 29 Power of this approach • Identified several functionally important genes and regions for conducting biological experiments
- 30. © 2022 Neo4j, Inc. All rights reserved. 30 Novelty of this approach • Provides a strong framework for systems biology paradigm in medicine • Patient-driven than patient-properties driven • Truly unbiased and integrative ◦ Not time dependent (based on scientific consensus) or subjective ◦ Correlation with survival combining multiple biomarkers has not been attempted
- 31. © 2022 Neo4j, Inc. All rights reserved. 31 Acknowledgements and support • Patients and their families – entrusting us with their data • Neo4j and Neo4j team (Rob Martin, Phani Dathar and Greg Shirley) • Yang Liu, PhD student – actively developing the database • Drs. Jason Huse, Vinay Puduvalli, Frederick Lang - unwavering support • Dr. David Jaffray, Sr. Vice President and CTO (MDACC) - insights, database availability • Moon ShotProgram® - coordinators and administrative support (neurosurgery, neuro-onc.) – funding and logistics to make the project possible • Brain Tumor Center, MD Anderson Cancer Center; The GLASS consortium • Rui Jiang, Suren – database support at MD Anderson Cancer Center • IT support staff at MD Anderson Dr. David Jaffray Dr. Jason Huse Yang, Liu
- 32. © 2022 Neo4j, Inc. All rights reserved. 32 Thank you! Questions?! Contact information: kskannan@mdanderson.org, kasthuri@gmail.com https://kannan-kasthuri.github.io/