Setting the Stage for Change The Practice of Pathology in the 21st Century: a Systems Approach José Costa, M.D. Department of Pathology, Yale School of Medicine RTICC Nov 08.
Sherlock Morgani Ackerman Babinsky Reubi Hamberger Stout Cabot Castelman Osler Foote Alzheimer Rudolph Rapaport Virchow, 1821- Laenec 1902 Popper Sir William Osler, 1849-1919 Pirani
Punctuated Change
Business Model Disruption in Health Care Hospitals become focused solution shops, practicing intuitive medicine Hypothesis Treatment Today s hospitals and specialist physician practices are agglomerations of solution shop, value-adding process, and (a few) facilitated network activities Focused value-adding process hospitals & clinics provide procedures after definitive diagnosis Facilitated networks take dominant role in the care of many chronic diseases 12/5/2008 Copyright Jason Hwang and Clayton M. Christensen 4
The Thesis Systems Pathology represents a practical adaptive response to the forces of change. Systems Pathology seeks to understand perturbed physiological systems and complex pathologies in their entirety by integrating all levels of functional and morphological information into a coherent model. It enables the design and testing of effective intervention and preventive measures. It is practiced by a combination of bottom-up data collection, often comprehensive (omics) and top down computational modeling and simulation.
Some Characteristics of Anatomical Pathology Interpretative Integrative Experiential
New Technologies for Molecular Analysis GGT wt GGT Protein Profiling Gene Mutation Arrays Gene Expression Arrays
Genomic Coverage of All Categories of CpG Islands Using Custom DNA Microarrays CLASS members probed % CpG islands in promoters Promoters without islands replicatio n subtotal 19,545 18,841 96.4 2.2 40,745 5,366 5,176 96.5 2.0 10,158 All promoter-associated 24,911 24,017 96.4 2.1 50,903 Unique, noncoding 43,033 41,587 96.6 1.9 78,237 CpGs at interspersed repeats 214,724 194,801 90.7 1.0 199,935 CpGs in tandem repeats 474 375 79.1 1.1 420 Subtotal 283,142 260,780 92.1 1.3 330,084 Total, other probes/controls 377,512
Hypomethylated Territories (example, chrom. 7) Recur in Multiple Tumors tumor margin normal
Methylome Profiles
Hierarchical Clustering Using Subset of Best 500 Methylation Probes
Scale of Organization
Segment, Classify, Feature Extraction Original image Segmented image Feature Statistics Classified image Aureon Labs.
Magic Developer Studio New Project
IF 20x: Epithelial Nuclei Network Red = 3 neighbors Aureon Labs
New Advances in Multiplexing Protein Detection CYTOKERATIN 14 CD 34 CYTOKERATIN 18 PTEN
Dynamic Range and Unmixing Parameters Total Gray Value vs. Dye Concentration with CRI Nuance 3.50E+09 Linear Response to Fluorescent Dyes Measurement Total Gray Value 3.00E+09 2.50E+09 2.00E+09 1.50E+09 Alexa 488 Alexa 555 Alexa 594 Alexa 647 1.00E+09 Dilution Series of Alexa Dyes 5.00E+08 0.00E+00 1 2 3 4 5 Dye Concentration 100:0 Starting Diltution for Alexa 555 and 568 (from original stock) Test Unmixing of Fluorescent Dyes with 100% Spatial Overlap 75:25 50:50 75:25 0:100 Series of Dye- Dye Ratios Measurement Software Processing 1.40E+08 1.20E+08 1.00E+08 8.00E+07 6.00E+07 4.00E+07 2.00E+07 0.00E+00 100:0 75:25 50:50 25:75 0:100 Alexa 555 Alexa 568 Corrected
KI-67 and AR Immuno-Detection System AR Ki-67
Cytoplam Segmentation and Classification MULTIPLEXED IMAGE EXTRACTED FEATURE NUMBER OF EPITHELIAL CELLS (DAPI+ CK18) AVERAGE SIGNAL INTENSITY IN THE CYTOPLASM INTENSITY DISTRIBUTION (%) 3+ 2+ 1+ Negative
Vessel Segmentation and Classification MULTIPLEXED IMAGE EXTRACTED FEATURE NUMBER OF LABELED VESSELS (CD34) TOTAL VESSEL AREA, PERIMETER, LENGTH, WIDTH MICROVESSEL AREA (AMVD) [ Lumen] INTENSITY DISTRIBUTION (%) 3+ 2+ 1+ Negative
CHARACTERIZATION OF THE ANDROGEN RECEPTOR ORIGINAL MULTIPLEXING SEGMENTED IMAGE DAPI+ (All Cells) CK18+ Epith Cells CK18+AR+ Epith Cells AR+ Stroma Cells Aureon Laboratories 2007)
CHARACTERIZATION OF THE ANDROGEN RECEPTOR AR Low Biochemical Recurrence (p=0.0001) AR High Time (months) Nl Epith Cells: CK18+AR+ Tm Cells: AR- Stroma CK18+AR+AMACR+ Cells Stroma Cells SEGMENTED IMAGE Aureon Laboratories 2007)
IF 20x: Ki-67 and pakt Aureon Labs
Biomarkers and Spatial Architecture Diversity in tissue units (nl or patho) Cell heterogeneity in BioM content Diversity in cell types Localization at the sub-cellular level
Prostate Needle Biopsy Studies COMPARING PROSTATECTOMY AND NEEDLE Bx, ABOUT 90% CORRELATION IN BIOMARKER DISTRIBUTION PRELIMINARY PREDICTIVE MODEL OF PSA RECURRENCE, CONCORDANCE INDEX OF 0.85
(Cordon-Cardo et al, J Clin Inves 117:370, 2007) PREDICTION OF TIME TO RECURRENCE AND RISK STRATIFICATION Probability of Remaining Recurrence-Free n = 262 P < 0.001 Validation - n = 682 Low-Risk (180 Pts., 171 Recurrence-Free) High-Risk (82 Pts., 54 Recurrence-Free) HR = 9 Actual Time to PSA Recurrence (months) PREDICTIVE ACCURACY: 86% - SPECIFICITY: 81%; SENSITIVITY: 85%
(Donovan et al, J Clin Oncol, In Press 2008) PREDICTION OF TIME TO CLINICAL FAILURE RISK GROUPS Probability of Remaining Clinical Failure-Free n = 345 P < 0.0001 Low-Risk (257 Pts., 252 Clinical Failure- Free) High-Risk (88 Pts., 63 Clinical Failure-Free) Validation - n = 385 HR = 11 Actual Time to Clinical Failure (months) PREDICTIVE ACCURACY: 92% - SPECIFICITY: 91%; SENSITIVITY: 90%
PREDICTION OF TIME TO CLINICAL FAILURE BIOPSY STUDY Green: AMACR Blue Nuclei: DAPI Red: AR Note: AMACR = Alpha-methylacyl-CoA racemase AR = Androgen Receptor Orange: AMACR(+) Epi Gland Green: AMACR(-) Epi Gland Blue: AR(-) / AMACR(-) Nuclei Red: AR(+) / AMACR(-) Nuclei Light Blue: AR(-) / AMACR(+) Nuclei Light Pink: AR(+) / AMACR(+) Nuclei Pink: Stromal Nuclei (Donovan et al, In Preparation 2008)
PREDICTION OF TIME TO CLINICAL FAILURE RISK GROUPS Proportion Clinical Failure Free n = 672 P < 0.001 Validation - n = 425 Low Risk HR = 3.5 High Risk Months Post-Prostatectomy (Donovan et al, In Preparation 2008)
How Tissue Analysis is Changing From subjective data gathering to objective data input. From a restricted modality of acquisition to integrating data across multiple scales. From microanatomy to functional morphology. From static data to dynamics. From monodisciplinary to transdisciplinary.
Eminence-Based Evidence-Based precision medicine
Multidisciplinary Advantage Mathematical Modeling Empirical Data In Silico Simulation Knowledge and Understanding
Montebello Model
Carcinogenesis is a microevolutionary process best described in the context of metapopulation dynamics Montebello Model of Tumor Formation J.Costa, Montebello 2005
Emerging Tumor Undergoing Disturbance
Metapopulation Dynamics with Disturbance Multi-species model with disturbance P i = cp(1 i D Σ i 1 P j ) e i P i Σ i 1 1 cpp i j Spatially explicit model of habitat destruction in a multi-species metapopulation model (Tilman, et al.)
The Montebello Model at simmering state
Interaction of Disturbance with mutation rate
Top Down (Network Topology) Dynamic Modification Combined Causal Action Complex Properties of the System Properties Shift Through Time Bottom Up (Molecular Structure) Variational Modification of Molecular Ensemble
Multidisciplinary Advantage Mathematical Modeling Empirical Data In Silico Simulation Knowledge and Understanding