Research

Digital Twins of Cancer Patients:

Since each cancer has its own unique characteristics, each one can respond differently to the same treatments. Therefore, the creation of a digital twin (DT) of cancer can assist us in predicting the evolution of an individual's cancer through modeling each tumor's characteristics and response to treatment. Hence, we take advantage of new advances in computational approaches and combine mechanistic, machine learning, and stochastic modeling approaches to create "My Virtual Cancer", a DT platform. This DT platform utilizes biological, biomedical, and EHR data sets. For each patient, the DT receives their information as input and predicts the evolution of their cancer.

In this project, we will focus on one common cancer type (breast cancer) and one rare cancer type (uveal melanoma) to evaluate the performance of the DT for both common and rare cancers.

Developing Data-Driven Models for Obtaining Personalized Cancer Treatments:

Quantitative and Systems Pharmacology (QSP) models, which are a system of differential equations modeling the dynamic interactions between drug(s) and a biological system, have been commonly used to discover, validate, and test drugs. These mathematical models provide an integrated “systems level” approach to determining mechanisms of action of drugs and finding new ways to alter complex cellular networks with mono or combination therapy to obtain effective treatments. Existing parameter estimation methods for QSP models often use assembled data from various sources rather than a single curated dataset. These datasets are usually obtained through various biological experiments, in vitro and in vivo animal studies.

We develop a framework to employ a combination of machine learning, mathematical, and statistical methods to improve available QSP modeling approaches so that they could be used for suggesting personalized cancer treatments. To obtain values of parameters of the QSP model for each patient separately, we first developed a TumorDecon software, which is a combination of recently developed methods, to predict the relative number of variables of the QSP model from patients’ data. We use patients’ data of primary tumor to estimate the values of parameters of the QSP model for each patient separately, instead of the common approach of assuming these parameters have the same values across all patients and using animal studies to estimate them. This new approach provides us with a unique opportunity to predict the efficacy of various treatments for each patient and suggest an effective personalized treatment strategy for cancer patients.

Exploring the Role of Transportation on Cancer Patients' Decision Making Through Machine Learning Techniques:

One of the main challenges of cancer patients is making decisions simultaneously about their cancer treatments and careers because of many factors, including side-effects and the cost of treatments. For example, the most common side-effect of cancer treatments is dizziness, which reduces the ability of patients in driving. This minor side effect might completely change cancer patients’ life, if the only way to get to work is driving. The main goal of this proposal is to investigate the role of transportation in decision making of cancer patients and their quality of life. To reach this goal, we have created a survey and we are analyzing its data by utilizing the recent advances in data science to identify the main factors that influence the quality of life of patients, and the relationship of these factors with each other. One of the main goals of this project is to make the process of decision making easier for patients and healthcare providers and help them to make better "scientific" decisions. The main goal of this proposed project is investigating the role of transportation in decision making of cancer patients and answer the following question. Is accessibility to the healthcare providers an important factor in the diagnosis or prognosis? Does transportation have any role in making decisions regarding treatments and occupations, i.e. choosing, changing, quitting a treatment or a job? Would existence of free/discounted rides from/to work or healthcare providers make any differences in cancer patients quality of life? This project seeks to investigate, through the lens of equity, the extent to which transportation mobility and healthcare accessibility influence various aspects of healthcare outcomes, including cancer diagnosis, testing, and treatment as well as treatment decision making and patients’ quality of life. We also investigate various mechanisms through which information and communication technologies could help cancer patients with a severe travel burden make better treatment decisions. We sincerely thank all patients participating in this study.

Finding Targeted Therapies for Patients With BAP1 Mutation:

In collaboration with Dr. Cebulla and Dr. Abdel-Rahman, we are exploring the role of BAP1 in carcinogenesis to find targeted therapies for patients with BAP1 mutation. BAP1 is a tumor suppressor gene important to the development and prognosis of many cancers, especially uveal melanoma (UM). We found that patients with UM and breast cancer, but not colon cancer, who died had a lower level of BAP1 gene expression compared to surviving patients. Importantly, in breast cancer patients, the lowest BAP1 expression levels corresponded to the dead young patients (age at diagnosis < 46). We also discovered that BAP1 is highly positively correlated with RBM15B and USP19 expression in invasive breast carcinoma, UM, and colon adenocarcinoma. All three genes are located in close proximity on the 3p21 tumor suppressor region that is commonly altered in many cancers. We also obtained cell toxicity and survival of 709 different cell lines to > 140 different drugs from the Genomics of Drug Sensitivity in Cancer database (http://www.cancerrxgene.org/). All drugs are either FDA approved or in late phases of clinical trials for treatment of various cancers. We also collected the BAP1 mutational and copy number variation status of the cell lines from the COSMIC cell line project and cancer the cell line encyclopedia. We assessed the relative effectiveness of drugs on cell lines with and without BAP1 mutation by multivariable analysis of variances. Three drugs showed statistically significant difference at a false discovery rate (FDR) of ≤ 25%. One of these three drugs was the PARP inhibitor rucaparib (effect size -0.87, pvalue=0.002, FDR 18%). Interestingly, another PARP inhibitor olaparib showed no significant selective effect (-0.13, pvalue=0.2, FDR 83%). The effects of these two PARP inhibitors were further validated in-vitro on BAP1 mutant and wild type cell lines. Studies of the effect of the two additional drugs are ongoing.

Modeling Cell Dynamics in Tumor Environments During and After Treatments:

Although the failure of cancer treatments has been mostly linked with the existence of resistant cells or cancer stem cells, new findings show a significant correlation between circulating inflammatory biomarkers and treatment failures. Most common cancer therapies such as surgery, radiation, and chemotherapy cause necrotic cell death, which activates the immune system in the same way as the wound-healing process. Therefore, we assume there is a wound that needs to be healed after stopping treatments.

The stochastic simulations of epithelial cell dynamics after a treatment, which only kills cells, show that higher fitness of cancer cells causes earlier relapses. Moreover, the tumor returns even if a single cancer cell with high fitness is involved in the wound healing process after such treatments. The model also indicates that the involvement of cancer cells in the wound healing after treatments leads to a fast relapse.

Additionally, cancer cells outside of the wound can cause a slow recurrence of the tumor. Therefore, the absence of relapse after such treatments implies a slow-developing tumor that might not reach an observable size in the patient’s lifetime. Conversely, a large solid tumor in a young patient suggests the presence of high fitness cancer cells and, therefore, a high likelihood of relapse after conventional therapies, which is consistent with clinical observations. The location of remaining cancer cells after treatments is a very important factor in the recurrence time. The fastest recurrence happens when high fitness cancer cells are located in the middle of the wound. However, the longest time to recurrence corresponds to cancer cells located outside of the wound’s boundary.

A New Hypothesis: Some Metastases Are The Result of Inflammatory Processes by Adapted Cells, Especially Adapted Immune Cells, at Sites of Inflammation:

We investigate the process of tumor initiation and progression in the sites of chronic inflammation. We argue that inflammatory carcinomas are initiated at the sites of chronic inflammation because chronic inflammation might cause cells to become adapted to the wound healing process. For example, immune cells might become adapted to send signals of proliferation and/or angiogenesis, and tissue cells might become adapted to proliferation (like the inactivation of tumor suppressor genes). By querying published available data sets, we calculated the probability of not observing any circulating tumor cells (CTCs) in blood from patients with various types of metastatic cancers. We determined that this probability is more than 1/3, i.e. at least one third of patients with metastatic cancer have no detectable CTCs in blood. Hence, there is another phenomenon besides CTCs in blood that causes metastasis. We propose that some metastases could be the result of the wound healing process by abnormal immune cells in inflammation sites. We propose that chronic inflammation induces the development of adapted abnormal immune cells, and adapted abnormal immune cells lead to the formation of the primary tumor because they send a high level of inflammatory signals such as proliferation signals, like IL-6. A new site of inflammation might recruit the adapted immune cells, in the same way these adapted cells will generate a tumor in this inflammation site. Note that immune cells are very active in the lung, liver, and bone, which are the common metastasis sites for inflammatory cancers. The existence of the pre-metastatic niche and metastasis to the site of injury also support this hypothesis. There is another possible mechanism: the new inflammation site recruits activated platelets. The activated platelets travel between sites of inflammation, which includes the site of inflammatory carcinoma. Tumor cells can link to adhesion receptors on platelets, so they can travel with activated platelets to the new site of inflammation. These activated platelets will start the wound healing process in this site, which now includes tumor cells. The tumor cells would respond to the wound healing signals more strongly than normal cells; thus a new tumor would initiate in this site of inflammation. If one of these hypotheses is true, then chemotherapy could facilitate the process of metastasis for inflammatory carcinomas. These hypotheses have several components including the chronic inflammation and cell adaptation, the wound healing process in treatments, the migration of immune cells between sites of inflammation, and the role of platelets. All these hypotheses need to be experimentally verified and computationally modeled. Testing these hypotheses is an interdisciplinary work: it needs computational models, some biological experiments, as well as patient data. We are eager to collaborate with experimentalists to test these hypotheses. We have also obtained some patient data to compare the results of treatments for inflammatory with non-inflammatory cancers. Additionally, we have obtained a protein expression data set for colitis-associated cancer, and we are using machine learning techniques to obtain the inflammatory networks.