Jackson K. Burton Abstract

Jackson K. Burton Abstract

 

Jackson K. Burton
  Ph.D. Candidate
  Applied Mathematics GIDP

  American Association for Cancer Research (AACR) Annual Meeting
  Pennsylvania, PA
  April 18-22, 2015

 

Abstract:

A model-based approach towards clinical pipeline optimization

Within a pharmaceutical company, the drug portfolio, often called the drug pipeline, is the company's collection of drugs in development.  Any one drug is categorized in one of the fives phases; pre-clinical, phase 1, phase 2, phase 3, and submission.  These phases are the development and testing phases as mandated by the FDA.  Most cancer drugs (about 90%) in development will fail to become approved by the FDA.  Due to the high financial risk of developing drugs that will eventually fail,  modeling approaches have been used to simulate drug pipelines over time in hopes of minimizing losses and maximizing profit. 

To analyze the progression of a drug pipeline, two theoretical modeling approaches are developed.  First a strictly mathematical approach called linear programming, (not in any way related to computer programming) is used to identify the  number of drugs that need to be in each phase in order to maximize  profit where the total drugs in the pipeline are restricted by the company's budget.  Company statistics  for the cost of developing the drug in a certain phase are used in this analysis.  The result the  indicates that the company should be very restrictive about what drugs are moved to the next phase as early as possible.  This analysis merely establishes what the optimal pipeline size for each phase should be, and thus does not describe the actual state of the company pipeline.

The second modeling approach uses a mathematical technique called a Monte Carlo process.  In this approach, the current company pipeline is simulated over time by advancing the drugs forward from phase to phase based on the company's success rates.  The model builds into it a stringency which puts greater restrictions on whether or not a drug advances.  The purpose of doing this was to quantify the effect of rejected bad quality drugs in early development thereby saving the company money by avoiding the cost of a failed drug later in phase.  Modeling such stringency was motivated by the linear programming result.  The fundamental result is that increasing stringency in the pre-clinical phase can increase profit dramatically. 

 

An in silico platform for characterizing ADC bystander effects

A relatively recent development in potential cancer treatment is the use of antibody drug conjugates (ADCs).  The proposed functionality of these drug molecules is as follows.  Highly toxic molecules are attached to antibodies via a molecular linker.  While linked to the antibody, these toxic molecules are non-destructive.  The antibodies themselves are designed to bind specifically to cancer cells, but not to healthy cells.  Upon binding, the cancer cell will be 'eat' the antibody, digest it, and thus release the attached toxic molecule killing the cell.  Once a cell is killed however, the once bound toxic molecules are now free to disperse and kill other neighboring cells.  This is what is defined as the bystander effect in this context.  Unfortunately, the bystander effect could be advantageous or detrimental depending on where the freely diffusing toxic molecule travels after the cell death. 

The promising use of these drugs is the high specificity to which a cancer cell can be targeted, but it is crucial to fully understand bystander effect in designing and optimizing ADCs.  Questions arise as to whether the bystander effect should be selected for or not.  In this work, a mathematical model was created to explore the bystander effect.  The model creates a 2-d spatial domain in which cancer cells and ADCs are planed randomly.  This is meant to mimic experimental settings in vitro where cancer cells are grown in shallow wells with ADCs added.  The model is then evolved forward in time and the resulting killed cells as well as the released toxin are tracked.

Because the model incorporates spatial structure, a distance metric can be defined to measure the bystander effect.  The functionality of the model allows for precise adjustment to a number of key parameters, such as the number of toxic molecules per antibody, the internalization rate (the rate at which the cancer cell eats the ADC), and the stability of the linker.  Simulations can be then be run to explore the design options that will optimize the bystander effect thereby optimizing treatment and minimizing the risk to the patient.