Data Science and Bioinformatics in Drug Discovery and Development
Updated: Jan 14
Introduction
When we think about Data Analysis or Data Science, the first thing that comes to our mind is its application in IT and Business. With all the advancements in technologies in the past decades, the use of data science are enormous. Data science and Bioinformatics tools play a critical role in drug development and discovery. The possibilities in which data science is used in biotechnology are amazing.
Drug discovery and development is a risky, complex, cost and time consuming process. Bioinformatics and data science made a positive effect on overall drug development process and biomedical research.
Definition
Data analysis, as the name states, is process of analyzing data. The data is collected, cleaned, transformed and interpreted to get useful insights. It plays a crucial role in preclinical drug development.
Data Science uses advanced math, statistical and predictive analysis, specialized programming, machine learning and artificial intelligence to make data driven decision-making and planning strategies.
Introduction to Bioinformatics
Bioinformatics is a diverse field that is a combination of computer science and biology, in which large and complex biological data is collected, stored and analyzed. By biological data, it means DNA and amino acid sequences. It uses big data (large datasets that cannot be processed with traditional tools and methods) for research purposes.
Human genome project
Human genome project is the mapping and sequencing of the human genome (entire sets of DNA), which took nearly 13 years to be completed. (1990-2003). This project provided the entire genetic map of humans and is a major breakthrough in drug development. It helps to research the genetic cause of any disease and enables the development of personalized and targeted medicine.
Data Collection
Data analysis plays an indispensable role in the early-stage drug development. The data for clinical research is collected either by clinical trials or by using real world data (RWD) which includes medical records, electronic health records, insurance claims, lab results and generated data from medical wearables (glucose monitors, smart watches etc.). RWD is cost efficient and less time consuming compared to traditional data collection methods.
Selecting Target Population
The suitable patient population is identified by various methods like selecting patients of interest with specific spectrum of disease or conditions, particular demography or backgrounds, using various scoring methods and comparing the different models.
With the availability of the human genetic map and new advanced tools, human genome sequencing can be done in a short amount of time and at a very low cost compared to the past few decades. This has made choosing suitable candidates very efficient.
Machine learning algorithms are used to analyze large clinical data and real world data to choose the target population by pattern recognition and make the trials efficient as possible.
Early detection of disease
Early detection and risk prediction of any disease or infection will help to treat early or prevent any new diseases.
Predictive modelling is used for analyzing the interaction between genes, proteins and other biomolecules. Artificial Intelligence and other machine learning algorithms like adaptive models, deep learning are also available, but they have their own challenges.
Biomarkers for disease and treatment
Biomarkers are biological molecules found in the body that act as an indicator to identify any normal or abnormal processes, that may be a sign of any disease, infection or effect of other environmental factor.
There are different functions of biomarkers such as, predicting disease or infection, knowing the status and the progress of the disease, showing the body’s response to the drug and treatment, monitoring exposure to risk or toxic substances due to use of medical treatments and chance of any recurrence of the disease. Biomarkers are even used in genetic level, to identify protein expressions and gene mutation.
Unlike traditional biomarkers which focus on single data types, AI algorithms can be used in large multi-dimensional biological data, preprocess the data, interpret and train the data using either classical machine learning model like regression model, random forest model, or deep learning. AI powered biomarkers have revolutionized drug development research.
Drug Design – In silico model
This technique is used to identify and design the drugs with computer models. It makes use of molecule modelling tools and computer aided drug designing methods (CADD). This helps in structure prediction and the optimization of the drug.
The major benefit of using this method is that it is highly cost efficient and saves more time and resources compared to, in vivo and in vitro trial and error methods. It makes early drug development very flexible. It also predicts potential for the drug, if there will be any toxicity or any other problems with the usage of the drug.
Some major challenges with in-silico models are their precision and effectiveness on the actual complex biological system. It also depends on the availability and quality of valid source data. After validation and decision making, in-silico modelling is followed by clinical trials.
Personalized and Optimized medicine
A vital advantage of personalized medicine is it improves the drug efficiency and optimizes the outcomes. It also reduces the negative effects of the drug. Recently protein drugs are developed to target specific protein targets depending on patient specific protein data.
Genome sequencing and analyzing helps in better understanding of the gene mutations, protein expressions and protein drug interactions with target site and molecules. This leads to personalized and optimized treatment, that can treat and even prevent diseases. This is a great revolution in drug industry that provides advanced human health.
Oncology
Machine learning is making a great breakthrough in the oncology drug discovery. Machine learning algorithms are trained to predict the compounds that bind to the specific target site, the toxicity and side effects of the drug for better understanding and decision making.
Real time imaging techniques like fluorescence and bioluminescence imaging are used to learn the behavior of tumor. Data Analytics is used to interpret data from these imaging, that helps in monitoring the interaction and effectiveness of the drug.
Post drug effects monitoring
Even after a drug passes the decision-making process and comes out to clinical trial, it is impossible to get all the information about the safety of the drug. Post marketing surveillance is necessary to monitor the effects of the drugs under various circumstances over the period.
Modern data mining that combines different tools and methods are used to collect adverse drug reaction (ADR) reports. More advanced design and statistical methods are required to analyze large data observations.
Challenges
There can never be any transformation and revolution without any challenges. There are some significant challenges and hurdles that might delay or hinder the drug development process.
First and foremost is the quantity and quality of the data from the early stages of drug development. This data is the primary requirement for understanding the disease or any condition.
The cost of the drug discovery process, and the time that involves the trial and approval process.
The strict protocols and regulations before decision making
The need for experts from many different disciplines like data scientists, biologists and healthcare professionals, to work together to model designs that align with real world standards.
Even a drug which shows promising results using AI might work differently when facing clinical trial and the chance being very low.
Conclusion
As experts work on AI biomarkers and models, the preclinical studies will become more precise and efficient and personalized. Future life science should embrace complex AI technology to translate complexity to actionable knowledge, leading to more effective treatment and therapy.
The concept of drug development using data science is all about understanding, what’s the right drug target - what right drug is against the right target - which patients will respond to the drug.
