Eugenia Floyd

Pathway analysis using random forests classification and regression

MOTIVATION Although numerous methods have been developed to better capture biological information from microarray data, commonly used single gene-based methods neglect interactions among genes and leave room for other novel approaches. For example, most classification and regression methods for microarray data are based on the whole set of genes and have not made use of pathway information. Pathway-based analysis in microarray studies may lead to more informative and relevant knowledge for biological researchers. RESULTS In this paper, we describe a pathway-based classification and regression method using Random Forests to analyze gene expression data. The proposed methods allow researchers to rank important pathways from externally available databases, discover important genes, find pathway-based outlying cases and make full use of a continuous outcome variable in the regression setting. We also compared Random Forests with other machine learning methods using several datasets and found that Random Forests classification error rates were either the lowest or the second-lowest. By combining pathway information and novel statistical methods, this procedure represents a promising computational strategy in dissecting pathways and can provide biological insight into the study of microarray data. AVAILABILITY Source code written in R is available from http://bioinformatics.med.yale.edu/pathway-analysis/rf.htm.

Tumor microvascular characterization using ultrasmall superparamagnetic iron oxide particles (USPIO) in an experimental breast cancer model

The diagnostic potential of ultrasmall superparamagnetic iron oxide particles (USPIO) for quantitative tumor microvessel characterization was assessed by kinetic analysis of dynamic magnetic resonance imaging (MRI) in a rodent breast cancer model. Microvascular characteristics (transendothelial permeability (KPS) and fractional plasma volume (fPV)) were estimated in 32 female Sprague Dawley rats, bearing breast tumors of varying malignancy. These values were compared to a prototype macromolecular contrast medium standard, albumin‐(GdDTPA)30. Transendothelial permeability (KPS) correlated significantly (P < 0.05) with the tumor grade (Scarff‐Bloom‐Richardson (SBR) score) for the USPIO (r = 0.36), as well as for the reference macromolecule, albumin‐(GdDTPA)30 (r = 0.54). Estimates for the fPV did not show a statistically significant correlation with the tumor grade for either contrast medium. In conclusion, USPIO‐enhanced MRI data were capable to characterize tumor microvessel properties in this breast cancer model: microvascular permeability (determined using USPIO) correlated significantly with tumor grade. Thus, quantitative estimation of microvascular characteristics in tumors could provide a surrogate of new vessel formation (angiogenesis) and thus a further important clinical indication for USPIO, in addition to MR angiography. J. Magn. Reson. Imaging 2001;13:882–888.

MRI assessment of microvascular characteristics in experimental breast tumors using a new blood pool contrast agent (MS-325) with correlations to histopathology

A new contrast medium, MS‐325, was compared to albumin‐(Gd‐DTPA)30 in 18 chemically induced rat breast tumors based on quantitative estimates of microvascular permeability (KPS) and fractional plasma volume (fPV) using a two‐compartment bidirectional model. No significant correlation was found between MS‐325‐enhanced microvascular assays with either tumor grade or with microvascular counts (MVCs). In comparison, the correlation coefficient between KPS and histologic tumor grade using albumin‐(Gd‐DTPA)30 (r = .58) was statistically significant (P < .01). Also, using albumin‐(Gd‐DTPA)30, a significant correlation (r = .55, P < .05) was observed between the KPS and MVC, a biomarker of angiogenesis. Correlations between fPV and MVC were not statistically significant for either contrast medium. In conclusion, using MS‐325, no significant correlations between the MR‐estimated permeability values or plasma volumes were observed in experimental breast tumors with either the histologic tumor grade or MVC. This analysis confirms our previous determination that capillary permeability estimates, using a prototype large molecular contrast medium, albumin‐(Gd‐DTPA)30, correlate significantly with both histologic tumor grade and MVC. J. Magn. Reson. Imaging 2001;14:237–242.

Development and Use of Biomarkers in Oncology Drug Development

Successful development and use of biomarkers will improve the productivity of oncology drug development. Recognition of the importance of biomarkers for speeding drug development is reflected in the precise definitions and concepts proposed by an NIH Working Group to standardize terminology and promote a more coherent and systematic approach to the development and use of biomarkers. Potential clinical biomarkers of drug efficacy are often identified through pre-clinical studies or basic research. Identification of potential biomarkers for use in oncology is moving rapidly forward through continuing advances in clinical imaging technologies, especially molecular and functional imaging. Other rapid advances are a product of the growing availability of new scientific reagents for established technologies and of high-throughput genomic and proteomic technologies that can generate hundreds of potential biomarkers for further evaluation. In certain cases, conventional clinical diagnostic techniques or assays can be adapted for use in pre-clinical models to evaluate their ability to serve as biomarkers for predicting clinical responses to new drug candidates. Evaluation (pre-clinical and clinical) of a potential biomarker is often the longest stage of biomarker development, and standards for evaluation or validation depend on the intended use and stage of clinical development. Biomarkers verified for use in preclinical studies can be used to help select appropriate animal models and lead compounds. Biomarkers verified for use in clinical trials can confirm a drugs pharmacological or biological mechanism of action, guide protocol design, aid patient and dose selection, and help to minimize safety risks. Oncology drug development can be optimized by using a tiered set of clinical biomarkers that predict compound efficacy and safety with increasing confidence at each rise in tier thereby aiding corporate decision-making about advancing compounds. In oncology, a special class of extensively evaluated biomarkers of efficacy (surrogate endpoints) that generally correlate with desired clinical outcomes can be used as a basis for corporate decisions as well as for gaining accelerated provisional regulatory approval of a drug.

Assessment of a rapid clearance blood pool MR contrast medium (P792) for assays of microvascular characteristics in experimental breast tumors with correlations to histopathology

The diagnostic potential of a new rapid clearance blood pool contrast medium (P792; MW = 6.47 kDa) for the MR assessment of microvessel characteristics was assessed in 42 chemically‐induced breast tumors, with comparisons to albumin‐(Gd‐DTPA). Microvessel characteristics, including the transendothelial permeability (KPS) and the fractional blood volume (fPV), were estimated by using dynamic MR data fit to a bidirectional two‐compartment model. The MR‐derived estimates for KPS and fPV using each contrast agent were compared, and assays using each contrast agent were correlated to the histologic tumor grade (SBR score) and the microvascular density (MVD) counts. Using P792‐enhanced data, neither KPS nor fPV showed a statistically significant correlation with the tumor grade or the MVD (P > .05). Conversely, using albumin‐(GdDTPA)30, KPS values correlated significantly with the histologic tumor grade (r = .55; P < .0005) and the MVD (r = .34, P < .05), whereas no correlation was established for fPV. In conclusion, based on P792 data no correlation between tumor microvascular characteristics and histologic markers (SBR score or MVD) was found in this breast tumor model. Our analysis suggests that contrast media of relatively large (on the order of 90 kDa) molecular size, such as albumin‐(GdDTPA)30, are more accurate for the characterization of tumor microvessels. Magn Reson Med 45:880–886, 2001.

The choice of region of interest measures in contrast-enhanced magnetic resonance image characterization of experimental breast tumors

Objectives:The objectives of this study were to determine if magnetic resonance (MR) estimates of quantitative tissue microvascular characteristics from regions of interest (ROI) limited to the tumor periphery provided a better correlation with tumor histologic grade than ROI defined for the whole tumor in cross-section. Methods:A metaanalysis was based on 98 quantitative MR image breast tumor characterizations acquired in 3 separate experimental studies using identical methods for tumor induction and contrast enhancement. Results:The endothelial transfer coefficient (KPS) of albumin (Gd-DTPA)30 from the tumor periphery correlated (r = 0.784) significantly more strongly (P < 0.001) with the pathologic tumor grade than KPS derived from the whole tumor (r = 0.604). KPS estimates, either from the tumor periphery or from the whole tumor, correlated significantly more strongly with histologic grade (P < 0.01) than MR image estimates of fractional plasma volume (fPV) from either tumor periphery (r = 0.368) or whole tumor (r = 0.323). Conclusions:KPS estimates from the tumor periphery were the best of these measurable MR image microvascular characteristics for predicting the histologic grade.

The anti-tumor activity of anti-CTLA-4 is mediated through its induction of IFNγ

Abstract The T-cell-specific receptor, CTLA-4, has been demonstrated to be a potent negative regulator of lymphocyte activation, the functional significance of which has been demonstrated in murine tumor models using blocking antibodies. However, the mechanism(s) involved in enhancing tumor regression has not been identified. In this study, we determined whether IFNγ was playing a role in this activity. In vitro, anti-CTLA-4 enhanced IFNγ production by lymph node cells obtained from tumor-bearing mice (351 pg/ml vs 77 pg/ml). Additionally, fibrosarcoma-challenged animals treated with anti-CTLA-4 had elevated levels of the IFN-inducible enzyme 2-5-OAS in draining lymph nodes (850 pM vs 260 pM for controls) and an increased amount of IFNγ in tumor lysates (at day 7, 620 pg/100 μg vs 160 pg/100 μg in controls). The importance of IFNγ was demonstrated by the ability of neutralizing antibodies to completely abrogate the anti-tumor effects of anti-CTLA-4. Moreover, fibrosarcoma cells were shown to be exquisitely sensitive to IFNγ-mediated class I upregulation and histological examination of tumors from anti-CTLA-4-treated mice revealed a trend toward increased tumor cell apoptosis and decreased angiogenesis. These studies have demonstrated that one mechanism for the anti-tumor effects of anti-CTLA-4 relates to its ability to augment IFNγ production, resulting in an increased expression of class I on the tumor, enhanced apoptosis, and a decrease in blood vessel growth.

Acute Drug-Induced Vascular Injury in Beagle Dogs: Pathology and Correlating Genomic Expression

Acute vascular injury that leads to vascular inflammation is a common finding in the preclinical toxicity testing of drugs in rats and dogs. However, the relevance of this finding for risk to humans is unclear. Concern about the safety of these drugs is heightened by the current lack of noninvasive clinical methods to predict the onset of vascular damage in animals or humans. Determining the relevance of this poorly understood preclinical outcome for humans requires a better understanding of the molecular mechanisms of injury in addition to the development of sensitive and specific leading biomarkers for the clinical diagnosis of acute vascular damage. Most molecular research on this toxicity has been performed in rats, but recent development of canine gene expression microarrays makes transcriptomic studies now possible in the dog. In this study, we investigated the molecular mechanisms of drug-induced vascular injury in dogs using gene arrays. After treating Beagles with toxic doses of CI-947, an adenosine receptor agonist, we profiled gene expression in the coronary arteries and correlated those changes with histopathology at 16 and 24 hours after dosing. The results demonstrated that pathobiological processes such as stimulation of the innate immune response, increased extracellular matrix turnover and oxidative stress were active at times of very early injury.

Dynamic MRI enhanced with Albumin-(Gd-DTPA)30 or ultrasmall superparamagnetic iron oxide particles (NC100150 injection) for the measurement of microvessel permeability in experimental breast tumors

Characteristics of the cell nuclei within human breast tumors, as assayed and scored by the Scarff-Bloom-Richardson (SBR) histopathologic method, correlate with patient prognosis and long-term survival rates (1,2). We have shown previously in rodent mammary tumors that the histopathologic tumor grade correlates with quantitative magnetic resonance imaging (MRI)–derived estimates of tumor microvascular permeability obtained using a prototypic macromolecular contrast medium, albumin–(GdDTPA)30 (3). Microvessel permeability depends on functional and morphologic characteristics of cancer vessels and on physicochemical properties of the injected solute/ contrast medium molecule (4–7). In this study, we focus on specific characteristics of two contrast media and their influence on MRI-derived estimates of microvessel characteristics. The increased permeability of tumor vessels to macromolecular solutes has been demonstrated repeatedly using albumin–(Gd-DTPA)30, which is used in this study as a reference standard (8–10). Although not previously tested as probes of endothelial hyperpermeability, ultrasmall superparamagnetic iron oxide (USPIO) particles, having a mean diameter of approximately 20 nm, were compared in this study to albumin–(Gd-DTPA)30, which is approximately 20% the size of USPIO. Unlike albumin–(Gd-DTPA)30, which is incompletely eliminated and potentially immunogenic, USPIO particles have a favorable pharmacologic and tolerance profile and are now being tested clinically (11). Thus, the purpose of this study was to define the potential of USPIO (represented by NC100150 Injection [Nycomed Imaging AS, Oslo, Norway]) for the quantitative characterization of tumor microvasculature, specifically for measures of the microvessel permeability and the plasma volume. Results were compared to estimates derived with the established macromolecular contrast medium, albumin–(Gd-DTPA)30, and to the histologic tumor grade.

The Trp53 Hemizygous Mouse in Pharmaceutical Development: Points to Consider for Pathologists

ILSI-HESI sponsored an international consortium for the evaluation of alternative models, including the Trp53+ /- mouse, for use in short-term carcinogenicity testing of pharmaceuticals. Products of the ILSI evaluation included guidance for protocol design and assay interpretation, spontaneous tumor incidences, diagnostic criteria for common proliferative lesions, and results of assays for pharmaceutical agents that are known human and/or rodent carcinogens and non-carcinogens. Based on the ILSI evaluation, recommended protocol elements for this model include: 26-week study duration, groups ≥15/sex/dose, a positive control group (benzene or p-cresidine), a negative control group and 3 dose groups, the high dose set at MTD or MFD, routine in-life evaluations, and complete necropsies with microscopic evaluation of tissues. Favored statistical analyses are trend tests or pair-wise comparisons, with no adjustments for survival. For an assay to be valid, positive control groups must demonstrate an effect, and the MTD or MFD must be reached in both sexes. Criteria for a negative response include a valid assay, no statistical increase in common tumors, no biologically significant numerical increase in rare tumors, and no tumor incidence above that of historical controls. Positive responses can consist of statistically significant increases in the incidence of a common tumor or numerical increases in a rare tumor, which may not be statistically significant. In either case, the incidence should be clearly above historical control values. Evidence of a dose response or occurrence of hyperplasia in a tissue with a neoplastic response can support interpreting an assay as positive. The two most common spontaneous tumors (>1%) in Trp53+/- mice are malignant thymic lymphomas and subcutaneous sarcomas. Use of implanted electronic transponders can increase the incidence of sarcomas. Important rare spontaneous tumors (incidence ≤1%) are osteosarcomas and pulmonary adenomas. Many other tumor types have been reported to occur sporadically in Trp53+/- mice. Diagnostic challenges for this model include differentiating lymphoma from atypical thymic hyperplasia and recognizing the variable histopathology of subcutaneous sarcomas. In reported bioassays, Trp53+/- mice responded positively to genotoxic carcinogens, negatively to non-genotoxic rodent carcinogens, and negatively to noncarcinogens, indicating that unlike the 2-year mouse assay, this short-term assay is not overly sensitive. Positive responses often elicited an increase in tumors that occur spontaneously. To successfully use this model, pathologists must understand the biology of the Trp53 tumor suppressor gene and the principles of protocol design and data interpretation for short-term bioassays. They must also know the historical response pattern of Trp53+/- mice to test agents and be able to accurately diagnose tumors in this model. Use of the Trp53+/- mouse presents the pharmaceutical industry with several challenges, one of which is managing the uncertainty created by a lack of precedents for regulatory decisions about some possible outcomes for short-term carcinogenicity assays.