Aydin Saribudak

Toward Genomic Based Personalized Mathematical Models for Breast Cancer Tumor Growth

Our Genomic Relevance Parameterization (GReP) model aims to explore a possible relationship between gene expression values from breast cancer patients and mathematical tumor growth modeling parameters calculated using data from clinical and preclinical measurements. We introduce two methods to relate genomic information and the tumor growth measurements. One method explores the impact of exponentiation of gene expression values, whereas the other utilizes the correlation between co-regulated genes and the growth parameters. As inputs to our GReP model, we used patient tumor volume measurements and genomic information for 74 breast cancer related genes from the I-SPY 1 TRIAL. We performed a preliminary validation of GReP model using experimental data from literature including MDA-MB-231 cell line, MDA-MB-231 cell line with CXCL12 gene over-expressed, and the MDA MB-231 sub-cell lines 1834 and 4175. Tumor growth curves generated by GReP model, for the initial exponential phase of tumor growth, closely match the pre-clinical data reported in the literature. These promising results show that it may be possible to build tools combining clinical information and genomic data to model cancerous tumor growth.

Genomic based personalized chemotherapy analysis to support decision systems for breast cancer

Personalized approach to anti-cancer therapy necessitates the adaptation of standardized guidelines for chemotherapy schedules to individual cancer patients. We introduce a methodology, namely Personalized Relevance Parameterization (PReP-G), based on the genomic data of breast cancer patients to compute time course of drug efficacy on tumor progression. The pharmacodynamic (PD) parameters of transit compartmental systems are computed to quantify the drug efficacy and kinetics of cell death. We integrate the genetic information of 74 breast cancer related genes for 78 patients with clinical t-stage of 3 from the I-SPY 1 TRIAL with the tumor volume measurements from NBIA database into our PReP-G model to compute tumor growth and shrinkage parameters. The performance of the method is evaluated for the breast cancer cell lines of BT-474, MDA-MB-435 and MDA-MB-231 for a given chemotherapy, where the anti-cancer agents Doxorubicin and Cyclophosphamide are administered to animal models and the change of tumor size is measured in time. We compare our results from PReP-G model with the experimental measurements. The consistency between computed results and the volume measurements is encouraging to develop personalized tumor growth models and decision support systems based on genetic data.

Mathematical models of tumor growth using Voronoi tessellations in pathology slides of kidney cancer

The impact of patient-specific spatial distribution features of cell nuclei on tumor growth characteristics was analyzed. Tumor tissues from kidney cancer patients were allowed to grow in mice to apply H&E staining and to measure tumor volume during preclinical phase of our study. Imaging the H&E stained slides under a digital light microscope, the morphological characteristics of nuclei positions were determined. Using artificial intelligence based techniques, Voronoi features were derived from diagrams, where cell nuclei were considered as distinct nodes. By identifying the effect of each Voronoi feature, tumor growth was expressed mathematically. Consistency between the computed growth curves and preclinical measurements indicates that the information obtained from the H&E slides can be used as biomarkers to build personalized mathematical models for tumor growth.

Quantification of cell apoptosis for in-vitro colorectal cancer cell cultures based on morphological features

Spatial properties of cell assays obtained by monitoring morphological characteristics of incubated cancer cells are studied as cell apoptosis indicators. Human colon carcinoma cells were cultured in 3d in-vitro microenvironment with FOLFOX treatment to experimentally evaluate cell viability for multiple days. With a 3D cell tracking algorithm guiding an inverted microscope in combination with a digital camera, bright field and fluorescent images of colorectal cells are captured at pre-determined time points. Time course of their spatial randomness properties based on the poisson point process and Voronoi features of cell apoptosis are computed from the fluorescent images. The results show that the heterogeneity among dead cells and the deviation among Voronoi polygon areas for cell apoptoses locations decrease as cell viability decreases. This relationship may be used as a biometric in drug efficacy measurements in in-vitro experiments.

Heterogeneity of colorectal cancer cell positions as a cell viability biometrie

In this paper, we quantify spatial heterogeneity in the positions of live cells as potential biomarkers for cell viability in 3d in-vitro microenvironment under the impact of anti-cancer agents. We present three case studies using human colorectal cancer cell lines of HCT-116, SW-620 and SW-480 in 48-hour experiments. With our data acquisition system guiding an inverted microscope and a digital camera, bright field and fluorescent images are captured at different time points. The viability of the human cancer cells, under FOLFOX impact are measured throughout the experiments. By processing the images captured from microenvironment, spatial heterogeneity in live cell positions are computed based on poisson point process. In all three case studies, we detect an increase in live cell spatial heterogeneity as the cell viability decreases in time as a result of FOLFOX. The encouraging result of correlation between cell viability and spatial heterogeneity has potential to be used in 3D in-vitro microenvironment assays as a biomarker.

A gene expression-based mathematical modeling approach for breast cancer tumor growth and shrinkage

In this paper, we introduce a personalized parameterization approach, namely prep-g, to explore impact of gene expression values from breast cancer patients on tumor growth and shrinkage characteristics using xenograft models. In construction of prep-g parameterization, in addition to individual effects of the breast cancer-related gene expressions, the impact of the correlation among them and the contribution of their multiple orders are considered. Tumor growth behavior, and delay and shrinkage effects of anti-cancer agents are examined in six case studies using xenograft models implanted with breast cancer cell lines. Tumor growth parameters for er+ cell lines bt-474 and mcf-7, and drug-related shrinkage parameters for cell lines mda-mb-231, mda-mb-468 and bt-474 under the monotherapy of drugs paclitaxel and doxorubicin are computed. Consistency of the experimental data reported in several studies in literature for multiple breast cancer cell lines in mice models and the computed results from prep-g are encouraging, which indicates that construction of mathematical models for tumor growth and shrinkage by combining gene expressions and clinical information may be feasible.

Spatial Heterogeneity Analysis in Evaluation of Cell Viability and Apoptosis for Colorectal Cancer Cells

In evaluation of cell viability and apoptosis, spatial heterogeneity is quantified for cancerous cells cultured in 3-D in vitro cell-based assays under the impact of anti-cancer agents. In 48-h experiments using human colorectal cancer cell lines of HCT-116, SW-620, and SW-480, incubated cells are divided into control and drug administered groups, to be grown in matrigel and FOLFOX solution, respectively. Our 3-D cell tracking and data acquisition system guiding an inverted microscope with a digital camera is utilized to capture bright field and fluorescent images of colorectal cancer cells at multiple time points. Identifying the locations of live and dead cells in captured images, spatial point process and Voronoi tessellation methods are applied to extract morphological features of in vitro cell-based assays. For the former method, spatial heterogeneity is quantified with the second-order functions of Poisson point process, whereas the deviation in the area of Voronoi polygons is computed for the latter. With both techniques, the results indicate that the spatial heterogeneity of live cell locations increases as the viability of in in vitro cell cultures decreases. On the other hand, a decrease is observed for the heterogeneity of dead cell locations with the decrease in cell viability. This relationship between morphological features of in vitro cell-based assays and cell viability can be used for drug efficacy measurements and utilized as a biomarker for 3-D in vitro microenvironment assays.In evaluation of cell viability and apoptosis, spatial heterogeneity is quantified for cancerous cells cultured in 3-D in vitro cell-based assays under the impact of anti-cancer agents. In 48-h experiments using human colorectal cancer cell lines of HCT-116, SW-620, and SW-480, incubated cells are divided into control and drug administered groups, to be grown in matrigel and FOLFOX solution, respectively. Our 3-D cell tracking and data acquisition system guiding an inverted microscope with a digital camera is utilized to capture bright field and fluorescent images of colorectal cancer cells at multiple time points. Identifying the locations of live and dead cells in captured images, spatial point process and Voronoi tessellation methods are applied to extract morphological features of in vitro cell-based assays. For the former method, spatial heterogeneity is quantified with the second-order functions of Poisson point process, whereas the deviation in the area of Voronoi polygons is computed for the latter. With both techniques, the results indicate that the spatial heterogeneity of live cell locations increases as the viability of in in vitro cell cultures decreases. On the other hand, a decrease is observed for the heterogeneity of dead cell locations with the decrease in cell viability. This relationship between morphological features of in vitro cell-based assays and cell viability can be used for drug efficacy measurements and utilized as a biomarker for 3-D in vitro microenvironment assays.

Gene Expression Based Computation Methods for Alzheimer's Disease Progression using Hippocampal Volume Loss and MMSE Scores

Personalized relevance parameterization methods (PReP-AD) based on artificial intelligence computation techniques are introduced to investigate the impact of gene expressions on Alzheimers disease (AD) progression. Our PReP-AD methods make use of the expressions of the genes that affect AD-related protein biomarkers (e.g., Aβ1-42 and tau proteins), mini mental state examination (MMSE) scores and hippocampal volume measurements from ADNI database for the patients with mild cognitive impairment (MCI), an intermediate stage from normal cognition to AD. For MCI patients, disease progression is computed with PReP-AD-MMSE and PReP-AD-HVL methods, where the former utilizes the change in MMSE scores and the latter based on the rate of hippocampal volume loss over time. The performance of both methods are assessed with an algorithm implemented using leave-one-out-cross-validation (LOOCV). The cognitive changes of AD patients with MCI stage are detected with both our MMSE score and hippocampal volume based computation methods. We observe an average error rate of 4.8% with PReP-AD-MMSE over a 72-month period and 1.63% with PReP-AD-HVL over 12 months. The promising results indicate that artificial intelligence based computation methods can be utilized to build decision support tools for AD progression.

Computation of Pharmacologic Therapy Effects on Cognitive Abilities of Alzheimer’s Disease Patients

To explore the impact of pharmacologic therapies on cognitive changes of Alzheimers disease (AD) patients, we develop an artificial intelligence (AI) based personalized relevance parameterization method, called PReP-AD-PH. Expressions of genes, which are effective in AD related protein biomarkers, and mini mental state examination (MMSE) scores of AD patients in mild cognitive impairment (MCI) stage are inputs for PReP-AD-PH. In this study, AD patients in MCI stage are split into two groups, such that the first group has 81 patients given monotherapy with cholinesterase inhibitor (ChEI) donepezil and the second with 70 patients received combinational therapy with donepezil and memantine. PReP-AD-PH computes parameters characterizing the cognitive changes in AD patients with MCI. Using a leave-one-out-cross-validation (LOOCV) based algorithm, we measure an average LOOCV error rate of 6.53% for patients received donepezil monotherapy, and 8.05% for those under combinational therapy. Cumulative distribution of LOOCV error rates of PReP-AD-PH results points out that AI based computation methods can be useful in assisting clinicians with pharmacologic therapy decisions for AD patients with MCI.