Can you fight cancer with the help of Big Data and Machine Learning? How can these technologies help in the procedure of diagnosis, drug discovery and treating cancer?

Cancer is an ailment with a long tail distribution. This implies there are different explanations for this condition to take place with no single solution to get rid of it. There are ailments which influence a huge number of people, however, have a sole reason for the event to occur. For instance, let us think about Cholera. Food or water tainted by Vibrio Cholerae is why Cholera occurs. Cholera can happen simply because of Vibrio Cholerae, and there is no other reason. When we discover the main source of an illness, it is moderately simple to overcome it. 

In our ammunition stockpile to pursue this war with cancer and overcoming it, Big Data and Machine Learning are weapons of mass destruction. 

Data Explosion and Gene Sequencing

One area that produces massive amount of data is gene sequencing. How much data? you may ask. Gene sequencing produces human data that is equivalent of ¼ of Youtube’s yearly data production. In terms of scale, this data combined with the additional information from genome sequencing if burned on 4GB DVDs, you would be looking at a stack that is half a mile high.

The strategies for gene sequencing have improved throughout the years, and the expense for the equivalent has plunged exponentially. In the year 2008, the expense on gene sequencing was 10 million dollars. Today, it works out to just a 1000 dollars. It is estimated to decrease further in the future. By 2025, it is estimated that 1 billion individuals will have gene sequencing done. It is evaluated that one billion individuals will have their qualities sequenced by 2025. By 2030, the genomics data will be somewhere close to 2 – 40 exabytes in a year. 

Fighting Cancer with Big Data and Machine Learning

The fight against cancer can be won in many ways if the large amount of data being generated is combined with Machine Learning algorithms. Diagnosis, treatment and prognosis assistance can be gained with Machine Learning. Customizable therapy will be possible, and the long tail distribution can also be dealt with.

Labelled data can be used in diagnosing cancer. This is made possible because of the vast Electronic Medical Records available and the data records from all hospitals. The use of Natural Language Processing is done to make sense of prescriptions of the doctors, CT and MRI scans are analyzed using Deep Learning Neural Networks. The various Machine Learning algorithms sift through the EMR database and find the patterns which are hidden. This will help with the diagnosis. 

An example to support this is a college student from the US was able to design a particular Artificial Neural Network from her home and even developed a model which was able to diagnose breast cancer with incredible accuracy.

Diagnosing with Big Data and Machine Learning

A very fine example of how diagnosis can be improved when it comes to cancer is the case of 16-year-old Brittany Wenger. She took it upon herself to improve diagnostics when her older cousin was diagnosed with breast cancer. To detect cancer, a less invasive method is FNA (Fine Needle Aspiration) which the doctor’s thought was not reliable. Brittany wanted to make this method better and decided to put her coding abilities to use to achieve this. An improved and less invasive method could be used by women if it was deemed reliable. 

Making use of the public domain data which was inclusive of FNA from the University of Wisconsin was the first step. An artificial neural network was then coded by her. Following this, she used cloud technologies for data processing and further trained the artificial neural network to detect similarities. It was a massive process of trial and error and she was finally able to detect breast cancer with FNA test data and it was sensitive to malignancy by 99.1%. This method isn’t restricted to breast cancer alone and is being used to detect other cancers as well.

The amount and quality of data determine the accuracy of the diagnosis. With more data available, database querying will be more by the algorithms. This will result in finding similarities and more valuable models being the output.

Treating Cancer with Big Data and Machine Learning

Moving on from diagnosis, Big Data and Machine Learning play a huge role in the treatment of cancer of as well. Let’s take another case where 49-year-old Kathy was diagnosed with stage III breast cancer. Kathy’s husband John was the CIO of a hospital in Boston. John planned Kathy’s treatment with the help of Big Data tools that were designed by him.

A powerful search tool was created in 2008 called SHRINE (Shared Health Research Information Network). This was created with the help of Harvard affiliated hospitals who shared their databases. By the time Kathy was diagnosed, the doctors treating her could sift through records that were close to 6 million in number. Questions like “Stage 3 breast cancer and treatment for 50-year-old woman” could be queried in SHRINE. This information allowed doctors to avoid surgery and treat her with chemotherapy drugs in customized treatment for the tumor cells.  

Once the chemotherapy was completed, the radiologists couldn’t find any more tumor cells in Kathy’s body. This is an example of how Big Data tools allow for customized treatment plan according to the patient’s requirement.

The one size fits all treatment process does not work for cancer because of its long tail distribution. For customized treatment plans to work, the following are key – diagnostic test results, gene sequence, gene mutation data, Big Data and Machine Learning tools.

Drug Discovery with Big Data and Machine Learning

Moving further from diagnosis and treatment, Big Data and Machine Learning can help revolutionize drug discovery. Open data and computational resources are used by researchers to discover new uses for drugs that are already in existence and have been approved by agencies like FDA. Another example of this was when a group of students at University of California. SFO used Big Data technologies and Machine Learning algorithms to find out that a drug used to treat pinworms could shrink a carcinoma which was a type of liver cancer in mice. This particular carcinoma was the second largest contributor of cancer deaths in the world.

Apart from finding new uses for drugs in existence, new drugs can also be discovered. Using data which is related to different drugs, their properties, chemical composition, disease symptoms, side effects etc, new drugs can be devised to treat various types of cancer. This will make it an easier process to devise new medicines and will help save millions of dollars in the process. 

In summary, Cancer is rather dangerous and comes in many different forms. In the form of Big Data and Machine Learning, we do now possess a stronger arsenal to combat cancer. From diagnosis, to treatment plans, to drug re-administering/discovery, we have the ability to beat cancer at every stage. 

Like this article? Subscribe to our weekly newsletter to never miss out!

Big data sets are so complex and large that common data processing tools and technologies cannot cope with them. The process of inspection of such data and uncovering patterns is called big data analytics. The basic question which arises in our mind is, “In what way is the drug discovery related to big data analytics?” or “how is data analytics useful in the process of drug discovery?”

The process of drug discovery requires the analysis, collection and processing of unstructured and structured biomedical data which is of large volume from surveys and experiments gathered by pharmaceutical companies, laboratories, hospitals or even social media. These huge amounts of data may also include data regarding sequencing and gene expression, molecular data which is included in drug data, data consisting of drug and protein interaction, data of electronic patient record and clinical trial, self-reporting and patient behaviour data in social media, data of regulatory monitoring, and literatures where protein-protein interaction and drug repurposing and trends may be found.

To examine in detail such diversified types of data in huge volumes to be able to discover new drugs, we need to have algorithms that are scalable, efficient, effective and simple. We now discuss how recent innovations in big data analytics improve the process of drug discovery. Algorithms are developed to uncover the patterns which are hidden in such data as unreported, discussions on drug side-effects in social media communications, sequencing and patient record data, drug-protein interaction and regulatory monitoring data, data regarding chemical-protein interactions etc., for the prediction of drug side-effects and how these types of predictions can be used to identify the possible drug structures with different necessary features. Big data analytics also contributes to much better drug efficiency and safety for regulators and pharmaceutical companies.

Upon implementing several measures of big data which are technology-enabled, pharmaceutical companies can enlarge the data they gather and enhance their approach to analysing and managing this data.

1.Integration of all the data

One of the biggest challenges facing the R&D organizations of pharmaceutical companies is having well-linked, consistent and reliable data. Data is the foundation upon which the value-adding analytics are built. Integration of efficient end-to-end data establishes an authoritative source for all the bits and pieces of information and correctly links different data which cannot be compared regardless of the source. Smart algorithms which link clinical and laboratory data, for example, could create automatic reports that identify applications or compounds that are related and raise red flags related to efficacy or safety.

2.Internal and external collaboration

R&D in pharmaceutical organizations is a secretive activity which is conducted within the R&D department with little external and internal collaboration. Pharmaceutical companies can extend their data networks and knowledge by enhancing their collaboration with external partners. Whereas end-to-end integration improves connecting the elements of data, the main aim of this collaboration is to improve the connections among all the stakeholders in delivery, commercialization, drug research and development.

3.Make use of IT-enabled portfolio for data-driven decision making

To make sure the allocation of scarce R&D funds is appropriate, it is critical to speed up decision making for pipeline and portfolio progression. Pharmaceutical organizations find it really challenging to make accurate decisions to about which assets to retain and which ones to kill. The financial or personnel investments they have made already may affect the decisions at the expense of merit and they lack decision-support tools which are appropriate to facilitate making calls which are tough. IT-enabled portfolio management enables the decisions which are data-driven to be made seamlessly and quickly. Smart visual dashboards must be used whenever there is a possibility to facilitate effective and rapid decision making.

3.Influence the new discovery technologies

Pharmaceutical R&D must continue using cutting-edge tools. These include systems biology and technologies that produce huge data very quickly. One of the examples for the technologies that produce huge data quickly is next-generation sequencing. This technology will make it possible to sequence an entire human genome within 18 to 24 months and at a cost of $100. The improved analytical techniques and wealth of new data will intensify the innovations of the future and feed the pipeline of drug development.

4.Deployment of devices and sensors

The advancement of instrumentation using miniaturized bio-sensors and the evolution of the latest smartphones and their applications are resulting in increasingly sophisticated health-measurement devices. Pharmaceutical companies are using smart devices to gather huge real-world data which was not available previously to scientists. Monitoring of patients remotely through devices and sensors constitutes an immense opportunity. This type of data can be used to analyse drug efficiency, facilitate R&D, create economic models which are new combining the provision of drugs and services and enhance future drug sales.

5.Raise the efficiency of clinical trials

A combination of smarter, new devices and exchange of fluid data will enable improvements in design of clinical trial and outcomes as well as higher efficiency. Clinical trials will become much highly adaptable to respond to drug-safety signals which are seen only in small but subpopulations of patients which are identifiable.

The following are the challenges facing transformation of bigdata in pharmaceutical R&D

1.Organization

The silos in an organization result in data silos. Functions usually have responsibility for their data and the systems they contain. Adopting a data-centric views, with a clear owner for each type of data through the data-life cycle and across functional silos, will greatly enhance the ability to share and use data.

2.Analytics and Technology

Pharmaceutical companies are following legacy systems containing disparate and heterogeneous data. These legacy systems have become a burden for these companies. Enhancing the efficiency to share data needs connecting and rationalizing these systems. There is also a scarcity of human resources supplied with a specific task of improving the analytics and technology needed to extract maximum value from existing data.

3.Mindsets

Many pharmaceutical organizations believe that unless they find a future state which is ideal, there is very less value to investing in enhancing the analytical capabilities of big data. Pharmaceutical organizations should gain knowledge from smaller, more entrepreneurial enterprises that see a lot of worth in the incremental improvements that get emerged from small-scale pilots.

Using Big data in pharmaceutical companies could slowly turn the tide of diminishing success rates and sluggish pipelines.

Conclusion

Effective utilization of big data opportunities can help pharmaceutical organizations better determine new ways to develop approved and effective medicines more quickly.

Like this article? Subscribe to our weekly newsletter to never miss out!

Follow @DataconomyMedia

Pande Lab at Stanford University in collaboration with Google released a paper earlier this week that focuses on how neural networks and deep learning technology could be crucial in improving the accuracy of determining which chemical compounds would be effective drug treatments for a variety of diseases.

A Google Research blog post explains how in the recent past, computational methods using deep learning with neural networks have attempted to ‘replace or augment the high-throughput screening process.’

So far, virtual drug screening has used existing data on studied diseases; but the volume of experimental drug screening data across many diseases continues to grow.

The paper titled “Massively Multitask Networks for Drug Discovery,” among other things, quantifies how the amount and diversity of screening data from a variety of diseases with very different biological processes can be used to improve the virtual drug screening predictions, explains the blog.

Working with a total of 37.8M data points across 259 distinct biological processes, using large-scale neural network training system to train at a scale 18x larger than previously used, the researchers managed to “probe the sensitivity of these models to a variety of changes in model structure and input data.”

“In the paper, we examine not just the performance of the model but why it performs well and what we can expect for similar models in the future. The data in the paper represents more than 50M total CPU hours.”

The entire effort, although it does not outline any milestone, is a step towards discerning an accurate and time saving method in drug discovery, that was traditionally almost impossible.

Follow @DataconomyMedia

(Image credit: Erich Ferdinand, via Flickr)