Sandeep Sanga

Mathematical modeling of cancer progression and response to chemotherapy

The complex, constantly evolving and multifaceted nature of cancer has made it difficult to identify unique molecular and pathophysiological signatures for each disease variant, consequently hindering development of effective therapies. Mathematical modeling and computer simulation are tools that can provide a robust framework to better understand cancer progression and response to chemotherapy. Successful therapeutic agents must overcome biological barriers occurring at multiple space and time scales and still reach targets at sufficient concentrations. A multiscale computer simulator founded on the integration of experimental data and mathematical models can provide valuable insights into these processes and establish a technology platform for analyzing the effectiveness of chemotherapeutic drugs, with the potential to cost-effectively and efficiently screen drug candidates during the drug-development process.

Predicting drug pharmacokinetics and effect in vascularized tumors using computer simulation

In this paper, we investigate the pharmacokinetics and effect of doxorubicin and cisplatin in vascularized tumors through two-dimensional simulations. We take into account especially vascular and morphological heterogeneity as well as cellular and lesion-level pharmacokinetic determinants like P-glycoprotein (Pgp) efflux and cell density. To do this we construct a multi-compartment PKPD model calibrated from published experimental data and simulate 2-h bolus administrations followed by 18-h drug washout. Our results show that lesion-scale drug and nutrient distribution may significantly impact therapeutic efficacy and should be considered as carefully as genetic determinants modulating, for example, the production of multidrug-resistance protein or topoisomerase II. We visualize and rigorously quantify distributions of nutrient, drug, and resulting cell inhibition. A main result is the existence of significant heterogeneity in all three, yielding poor inhibition in a large fraction of the lesion, and commensurately increased serum drug concentration necessary for an average 50% inhibition throughout the lesion (the IC50 concentration). For doxorubicin the effect of hypoxia and hypoglycemia (“nutrient effect”) is isolated and shown to further increase cell inhibition heterogeneity and double the IC50, both undesirable. We also show how the therapeutic effectiveness of doxorubicin penetration therapy depends upon other determinants affecting drug distribution, such as cellular efflux and density, offering some insight into the conditions under which otherwise promising therapies may fail and, more importantly, when they will succeed. Cisplatin is used as a contrast to doxorubicin since both published experimental data and our simulations indicate its lesion distribution is more uniform than that of doxorubicin. Because of this some of the complexity in predicting its therapeutic efficacy is mitigated. Using this advantage, we show results suggesting that in vitro monolayer assays using this drug may more accurately predict in vivo performance than for drugs like doxorubicin. The nonlinear interaction among various determinants representing cell and lesion phenotype as well as therapeutic strategies is a unifying theme of our results. Throughout it can be appreciated that macroscopic environmental conditions, notably drug and nutrient distributions, give rise to considerable variation in lesion response, hence clinical resistance. Moreover, the synergy or antagonism of combined therapeutic strategies depends heavily upon this environment.

Nonlinear modeling and simulation of tumor growth

The effects of the interaction between cellularand tumor-scale processes on cancer progression and treatment response remain poorly understood (for instance, the crucial role of the microenvironment in cancer growth and invasion [95, 65, 184, 183, 160, 110, 66, 166]). Three-dimensional tissue morphology, cell phenotype, and molecular phenomena are intricately coupled; they influence cancer invasion potential by controlling tumor-cell proliferation and migration [78, 187, 198]. Hypoxia [88, 210, 186, 70, 91], acidosis [91, 199, 96], and associated diffusion gradients, caused by heterogeneous delivery of oxygen and nutrients and removal of metabolites [104, 103] due to highly disorganized microvasculature [92, 106] and often exacerbated by therapy (e.g., anti-angiogenic [160, 177]), can induce heterogeneous spatial distribution and invasiveness of tumor cells through a variety of molecular [175, 209, 208, 44, 165, 173, 145, 118, 28, 29, 190, 156, 185, 120, 122, 24, 160, 177, 174, 61] and tissue-scale [59, 127, 72] mechanisms corresponding to different tissue-scale invasive patterns [78, 164, 194, 167, 111, 187, 198, 204, 117, 206, 64, 179, 75, 77, 53, 172]. Such complex systems, dominated by large numbers of processes and highly nonlinear dynamics, are difficult to approach by experimental methods alone and can typically be better understood only by using appropriate mathematical models and sophisticated computer simulations, complementary to laboratory and clinical observations. Mathematical modeling has the potential

Using mathematical models to understand the time dependence of the growth of ductal carcinoma in situ.

Abstract #1165 Background: Models of cancer growth have been developed that predict tumor size and growth dynamics for invasive tumors. However, it has been difficult to model ductal carcinoma in situ (DCIS) because of the constraints introduced by its containment within the duct system.
 Materials and Methods: We have developed a spherical model of growth of solid type DCIS using chemical engineering models of reaction and diffusion in porous media to represent the spread of DCIS in the duct systems. The model predicts tumor diameter based on four input parameters: the ratio of the apoptosis rate to the proliferation rate (A), the diffusion penetration length for nutrient to sustain the tumor growth (L), the volume fraction that tumor cells occupied within the involved breast tissue (V), and the time taken for a cell to complete mitosis(T). We have estimated L, V, and T from the literature, and then back-calcuated A for a range of diameters. We have used these four parameters as inputs and studied the time dependence of the evolution of DCIS.
 Results: We have found that the range of the values of A that we determined are within an adeqaute physiological range based on rates of proliferation and apoptosis taken from the literature. Using the model, the time to reach at least 95% of the maximum size ranges from less than 30 days for DCIS measuring 0.5 cm to almost 80 days for DCIS measuring 6 cm in diameter.
 Discussion: There has been little understanding of how long it takes for DCIS to grow, and whether it reaches a steady state size. Our simulations show that DCIS can grow to sizes as large as 6 cm in less than 3 months if it has the correct properties, including a high proliferation rate relative to the apoptosis rate and appropriate access to nutrients. This finding may help to explain why many cases of DCIS are not diagnosed before they progress to invasive carcinoma. Citation Information: Cancer Res 2009;69(2 Suppl):Abstract nr 1165.

Abstract 4870: GenePool: A cloud-based technology for rapidly data mining large-scale, patient-derived cancer genomic cohorts including The Cancer Genome Atlas

Proceedings: AACR 106th Annual Meeting 2015; April 18-22, 2015; Philadelphia, PA GenePool is a cloud-based system for the secure storage, management, analysis, visualization and interpretation of large-scale human genomics data. The GenePool platform enables rapid interrogation of massive amounts of data generated by current laboratory methods and was designed to meet the needs of scientists engaged in cancer biomarker discovery and characterization. The software solution provides users with an intuitive interface, an easy way to store and select genomics datasets according to sample-associated metadata for analysis, and an automated system for performing routine, well-characterized genomics workflows. Results are presented with multiple options for annotation, sorting and data export, with a visualization tool that facilitates browsing of genomic data for biomarker identification. We will demonstrate how GenePool removes the data download, management, and computing burdens faced by researchers interested in working with large amounts of patient-derived cancer genomics exemplified by The Cancer Genome Atlas. To date, TCGA has completed molecular characterization of more than 9,000 patient samples from a target of 17,000, and will have released an estimated 2.5 petabytes of data at project completion. We will present case studies using the >25 cancer cohorts worth of RNA-Seq, miRNA-Seq, Exome Sequencing, Protein Expression, Copy Number, and DNA Methylation data (provided to the community through the TCGA Data Portal) and made readily available for sample selection and analysis within GenePool. As validation of the GenePool platform, we will present results obtained within minutes that validate knowledge reported in the literature over the course of decades. We will then present novel findings for less understood cancer indications. Note: This abstract was not presented at the meeting. Citation Format: Sandeep Sanga, Antoaneta Vladimirova, Richard D. Goold, Tod M. Klingler. GenePool: A cloud-based technology for rapidly data mining large-scale, patient-derived cancer genomic cohorts including The Cancer Genome Atlas. [abstract]. In: Proceedings of the 106th Annual Meeting of the American Association for Cancer Research; 2015 Apr 18-22; Philadelphia, PA. Philadelphia (PA): AACR; Cancer Res 2015;75(15 Suppl):Abstract nr 4870. doi:10.1158/1538-7445.AM2015-4870

Abstract 4279: Rapid biomarker discovery using large-scale, patient-derived cancer genomic cohorts

Target discovery and validation in oncology has largely relied on molecular and functional studies performed in cell lines. Recent advances in genomics have now created large databases based on well-characterized tumor tissue, which has enabled direct investigation of patient tumors for novel targets and predictive markers. Following these discoveries, it is routine to perform functional studies in cell line-based systems; however, it is often challenging to find a relevant cell line model and if found, there are often numerous factors which confound biology when using historical cell lines for functional studies. The result can be a process, which takes considerable time and does not readily translate to clinical relevance. We report here our efforts to identify novel targets and markers through data mining large cohorts of cancer genomics data consisting of samples deriving from both drug-naive and -treated patient-derived cancer samples. To serve as the drug-naive reference, we look to the thousands of patient-derived tumor specimens, covering 30+ tissue types, genomically characterized by The Cancer Genome Atlas. For both drug-naive and drug-treated primary data, we leverage Molecular Response9s proprietary bank of viable cryopreserved tumor cells. The bank contains more than 144,000 tumor specimens, covering 25 tissue types and 76 clinical diagnoses. Each of the specimens has been profiled against a panel of drugs ex vivo, and levels of resistance recorded. Nearly 400 tumor specimens from Molecular Response9s proprietary bank have been genomically characterized. Station X has designed and developed a software environment called GenePool for the management, analysis and communication of genomic information for cohort-scale biomarker studies. The genomics data along with sample-associated clinical metadata deriving from Molecular Response9s bank and The Cancer Genome Atlas were imported into GenePool. We evaluated consistency between drug-naive samples deriving from the Molecular Response bank and TCGA, and subsequently mined for expression-based markers differentiating varying grades of drug sensitivities from drug-naive specimens. We are currently evaluating these markers for further functional study in patient-derived models and plan to report to these findings. Citation Format: Sandeep Sanga, Praveen Nair, Cyrus Mirsaidi, Thomas Broudy. Rapid biomarker discovery using large-scale, patient-derived cancer genomic cohorts. [abstract]. In: Proceedings of the 105th Annual Meeting of the American Association for Cancer Research; 2014 Apr 5-9; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2014;74(19 Suppl):Abstract nr 4279. doi:10.1158/1538-7445.AM2014-4279

Abstract 1138: Rapid validation of a novel kinase target using a large scale genomic database and matched patient derived tumor models.

Proceedings: AACR 104th Annual Meeting 2013; Apr 6-10, 2013; Washington, DC Target discovery and validation in oncology has largely relied on molecular and functional studies performed in cell lines. Recent advances in genomics have now afforded large databases based on well-characterized tumor tissue, which has enabled direct investigation of patient tumors for novel targets. Following these discoveries, it is routine to perform functional studies in cell line-based systems; however, it is often difficult to find relevant cell line model and if identified, they can be confounded with artifacts created from decades of culture. The result can be a target validation process which takes considerable time and does not readily translate to clinical relevance. We report here the creation and use of a target validation platform based on a large-scale genomic database matched to patient-derived tumor models. The platform relies on Molecular Responses proprietary bank of more than 144,000 patient derived tumors, of which nearly 400 tumors have been genomically characterized and databased for target discovery studies. The database is growing, but currently features the following cancer indications: colon carcinoma, NSCLC, melanoma, ovarian carcinoma, prostate and non-hodgkins lymphoma. Upon discovery of a novel target, tumors of interest are immediately implanted into mice to perform functional studies in direct patient derived models–either in vivo or using mouse-passaged cells for broader ex vivo studies. Through use of this platform, we have identified a novel kinase target for potential therapeutic development. We investigated prevalence of target overexpression across 7 cancer indications, and identified melanoma as a clinical indication of high interest. We examined growth characteristics from patient tumors featuring high kinase gene expression vs. low expression to help characterize the role of this target in oncology disease progression. Finally, we performed functional knockdown studies in patient derived models to further validate this novel kinase as a druggable target of pharmaceutical interest. Studies are ongoing to develop small molecule and antibody-based therapeutics that will serve as drug candidates for further development. Citation Format: Thomas Broudy, Sandeep Sanga, Jill Ricono, Mohit Trikha, Cyrus Mirsaidi, Kesavan Nair Praveen. Rapid validation of a novel kinase target using a large scale genomic database and matched patient derived tumor models. [abstract]. In: Proceedings of the 104th Annual Meeting of the American Association for Cancer Research; 2013 Apr 6-10; Washington, DC. Philadelphia (PA): AACR; Cancer Res 2013;73(8 Suppl):Abstract nr 1138. doi:10.1158/1538-7445.AM2013-1138

Neural Network with K-Means Clustering via PCA for Gene Expression Profile Analysis

Gene expression microarray data are highly multidimensional and contain high level of noise. Most of these data involve multiple heterogeneous dynamic patterns depending on disease under study. In addition, possible errors might also be introduced along data collection path if multiple sites and methods are used. In this paper a combined data mining method, i.e., neural network with K-means clustering via principal component analysis (PCA), is proposed to address the data complexity issues when conducting gene expression profile mining. The proposed method was tested on gene expression profile in lung adenocarcinoma, collected from multiple cancer research centers, for survival prediction and risk assessment. The results from the proposed method were analyzed, and further studies for future improvement of the proposed method were also recommended