What is Machine Learning? (And Why Use It?)

,

What is Machine Learning?

The science of machine learning isn’t new, but interest in the field is gaining fresh momentum. The concept of technology allowing machines to learn from data dates to the 1950s, but new applications of machine learning have the potential to change the face of pharma forever.

But what is it exactly, and how is it being used currently? SAS defines machine learning as:

“A method of data analysis that automates analytical model building. Using algorithms that iteratively learn from data, machine learning allows computers to find hidden insights without being explicitly programmed where to look.”

All machine learning relies on data – and the healthcare industry is a data goldmine. Big data and machine learning in pharma could generate an annual value of $100 billion across the US, according to new research from McKinsey.

New drugs cost $2.6 billion to develop and take an average of 12-14 years to get to market

New drugs cost $2.6 billion to develop and take an average of 12-14 years to get to market, meaning the industry has a major challenge on its hands. Machine learning has the potential to reduce these costs, enhance competition, accelerate drug discovery, boost quality compliance, and cut time to market.

Let’s look at just three examples of how machine learning is helping the industry to evolve.

Improving patient safety and drug development

IBM Research has an impressive track-record of applying machine learning to medicine. In April, the company received a patent for a machine learning tool that can identify the side effects of certain therapies and diseases. This data can be used to improve patient safety and identify new applications for drugs, as well as improve the efficiency and effectiveness of drug discovery and development.

The tool has already established a link between weight loss and mood disorders such as bipolar and depression. This could lead to researchers examining ways of repurposing mood-disorder drugs to be used as a form of weight control.

Driving value-based, precision medicine

Celgene is a company committed to accelerating value-based, precision medicine development through machine learning. The biopharma giant has investments in a machine-learning platform to support decisions across multiple therapeutic areas.

As a result, Celgene can apply this technology to drug discovery, clinical development and commercialization. This means greater insight and answers to more complex questions for treatments when matching drugs and other health interventions to the individual.

Unlocking the secrets of stem cells

The team at the Allen Institute for Cell Science launched a vast online catalogue of 3D stem cell visuals. Each image in the Allen Cell Explorer has been created through deep learning analysis. Computer scientists used the technology to establish relationships between the locations of structures within cells and then create predictive models of cell organization. The program ‘learns’ by comparing its predictions to actual cells.

In the future, the portal will include distinct cell types and could help researchers identify stem cell lines with mutations linked to cancer and other diseases.

Turning content into smart data

Life sciences organizations are also increasingly turning to machine learning to extract insights from published scientific research. And with good reason – this content is the lifeblood of the organization, and its volume and velocity are growing. More scientific research published between 2010 and 2014 was indexed in MEDLINE than all research published before 1970.

However, these initiatives must surmount challenges to content licensing and data preparation – including lack of access to full-text scientific literature vs. abstracts, inconsistent terms of use, and formatting discrepancies. Informatics teams, data scientists, and knowledge managers must also make schema and template decisions that will affect the types of features and attributes that can be extracted from the content. How content is processed into smart data upstream affects results from machine learning processes downstream.

These challenges aren’t trivial, but in the competitive life sciences market, breakthroughs in these areas could produce big outcomes in cost and time to market.

Related Reading: Accelerating Drug R&D While Minimizing Costs

Doug Knight

Author: Doug Knight

Doug has a Software Engineering background in a number of different industries, including Weather, Medical, NASA, Email Marketing, and Defense. From his early years on Hubble Space Telescope to more recent work at MITRE, and his previous position as Development Manager at Net Atlantic, Doug’s knowledge spans software development, project management, team building, and system architecture disciplines. He is Technical Lead for the XML for Mining and eCommerce projects at CCC. He is a current 10 year resident of Marblehead Massachusetts, with family spread between here, Maryland, and Virginia. He has a daughter age 15 who looks to be following in her Dad’s footsteps, already interested in programming and digital design and art. Doug’s hobbies include reading (fiction mostly), tinkering with computers and networks, and meditation.

For inquiries related to this blog, email: sweston@copyright.com