Machine Learning Application: Organs-on-a- chip in Parellel

Abstract

Cancer is one of the most common deadly diseases in the world. This is due to the fact that no good cure is available. Many medications developed for this disease are studied in vitro and then tested in animals (such as rats, pigs, sheep, and in some cases primates). Most of this testing prove successful but may fail to do so in clinical trials (in human body). The failure of the drug testing during the clinical trial is a result of the dissimilarity between animals and human. By developing a human tissue or organ-on-a-chip, the structure, physiology and functions of a human organ can be represented on a chip. When many of these human organs are integrated in parallel on-a-chip-devices, a high throughput screening system for drug discovery is obtained. However, this will involve the production of large amount of data which will require artificial intelligence, neural networks or machine learning data analytic methods to analyze and derive reliable conclusions. Produced data consist of the numeric values for the parameters necessary for testing drug efficacy (such as the solubility of drug and iron levels in the blood stream) that are measured from the integrated chip. Machine learning is the most affordable and reliable data analytic method which learns from a training data in order to predict or classify an output based on a similar test data. The training data in this case are the exact numerical values of the parameters necessary for determining the drug efficacy with their expected cell reactions which were defined after the drug was developed while the test data are the measured numerical values of the parameters from the integrated chip whose cells reactions to the drug need to be analyzed. This thesis therefore aim at the outlining the present method used for testing the efficacy of developed drugs, the reasons why these drugs fail in clinical trial and the use of organs–on-achip-in parallel as an improved method for the drug development process. Data is generated with respect to the values of parameters used in determining the efficacy of a drug such as Cisplatin (a drug for treating cancer) and a support vector machine learning algorithm is trained to test these parameters in order to classify and predict the reactions of cells. This will give rise to the claim that if multiple human organs are developed on-a-chip and integrated on-a-chip-devices, the integrated devices will act like a human body and when contaminated with a disease, a developed drug can be scan through the integrated chip to identify the reactions of the cells to the drug thereby determining the drug efficacy. This will reduce the time spent in testing drugs as well as the risk of mortality during clinical trials or even after drug approval and use.