Now Reading
Chennai-based AI Drug Discovery Startup Vingyani Uses ML Algorithms For Toxicity Prediction

Chennai-based AI Drug Discovery Startup Vingyani Uses ML Algorithms For Toxicity Prediction

Srishti Deoras

Download our Mobile App

Chennai-based Vingyani is probably one of the few data science companies that is supporting the data and analytical needs of biopharma industry. Founded in 2017 and led by Srinivasan Parthiban, the team has implemented machine learning algorithms for toxicity prediction observing some preliminary results. Motivated and inspired by this, they are working on prototypes to bring the most definitive results.

Vingyani uses biosciences domain knowledge to solve complex and novel business problems with data-driven systems. “Our capabilities start with extraction, normalisation or standardisation and integration of larger data”, shares Parthiban.

Analytics India Magazine got in touch with Parthiban, who shares his interesting journey in biosciences and artificial intelligence, AI in drug discovery, it’s scope in India and more.

Story Behind Vingyani

After a PhD from Indian Institute of Science, and a postdoc from NASA Ames and Weizmann Institute,and stints at AstraZeneca, Jubilant Biosys, GvkBio, Parthiban took a break and travelled to USA and Europe to study the emerging area. AI and ML were the buzzwords at the moment, and the idea of Vingyani came from the realisation that his academic research work on mathematical modelling is related to the current machine learning technologies, and that his industrial experience on data curation is the ideal foil to mathematical modelling to succeed in AI.

AI And Advanced Analytics In Pharma

Parthiban believes that the promise of AI has hit the biopharmaceutical industry in full force. “Using these transformative innovations, we plan to collect and use a range of very large data sets to train ML models that will help address key problems in the drug discovery and development process”, he shares. To ensure this, his team of AI researchers and scientists leverage high-quality data which has been collected explicitly with machine learning in mind from the very start.

Explaining the need to have AI and advanced analytics in pharma, Parthiban shares that since biomedical knowledge is huge and keeps growing exponentially, there are millions of new documents every year. “While our human brains are limited in making sense out of all this, machines are not”, he said. And therefore they are designing an AI platform  that will be a science assistant and that they hope will learn overtime and become a scientist herself.

Also, biopharmaceutical companies are investing in new data and analytical capabilities. With a large amount of data in the form of biological measurements corresponding to patients and samples, applying statistical methods used for low dimensional data might give wrong results. This is known as the curse of dimensionality problem, when there are a lot more covariates than samples (short and wide data). The team is trying to solve this with their analytics solutions.

AI And Drug Discovery In India

Vingyani is also exploring big data and machine learning in drug discovery to make processes faster, cheaper, and (most importantly) more successful. “At Vingyani, our vision is to transform drug discovery from a slow, sequential, and high-failure process into a rapid, integrated, and patient-centric model by integrating high-performance computing, machine learning methods and bio activity data”, he said.  

As Parthiban explains, discovering a new drug typically involves following steps:

  • Identify a medical need
  • Identify the underlying biological mechanism
  • Design a molecule acting via this mechanism to produce the desired effects

“Though is may sound simple, in truth it isn’t. Leaving aside biology, an important part of the problem is the chemistry: there are just too many molecules to choose from, perhaps as many as 1060 for all drug-like small molecules. Even with the help of computers one cannot enumerate more than a few billions of them, let alone predict their possible biological activity or how to synthesize them”, he said.

He plans to deal with the situation by asking these questions — can we automate the process and take the manual design out of the equation? Parthiban believes that this is possible using deep learning.

Has India arrived there yet?

Parthiban says that though Bioinformatics has not sustained as a separate industry, it evolved into Indian Contract Research Organizations later. “We see a similar scenario now and the time is good for investing in PharmaAI and the front runners will have a great opportunity to grow big in India”, he said.

“Discovering a drug costs $2.6 billion and it takes 15 years to bring it to market. Drug makers need to find a more efficient way of developing medicines. Big breakthroughs happen when what is suddenly possible meets what is desperately necessary.  The biopharmaceutical industry is experiencing a fundamental shift in how data is used to drive business insights”, he added.

See Also

Stagnant productivity, an increased pressure to remain competitive, the shift from volume to value and patient-centric care are forcing biopharmaceutical companies to evolve how they generate and use data.

Technology Stack At Vingyani

Parthiban shares that their AI stack consists of two components — infrastructure and developer environment.   

  • Infrastructure refers to the tools, platforms and techniques used to run and store data, build and train AI algorithms. They use GPUs and AWS for raw computational power to run AI algorithms. They also use TensorFlow, Keras and PyTorch as primary machine learning platforms, along with several libraries for mathematical operations.
  • Developer environment refers to the tools that assist in developing code to bring out AI capabilities. The workflow automation tools include Anaconda and GitHub. Jupyter Notebook is also integrated into development environment. Visualization choices include MatPlotLib, Seaborn and D3.JS.  

Challenges On The Way

Parthiban lists few challenges as below:

  • The first challenge is choice of the problem that we address and market it. Since the AI companies are growing at a breakneck speed, we need to keep track of the new technologies. Otherwise, by the time, we build the product and go to the market, a superior quality product will be available for free download.
  • Retaining the talents is another challenge.            
  • Last but not the least, the start up companies always look for various grants schemes from the government. It is important to note that the process takes more than six months to know the outcome of the review. Given this, having a sound financial background is important as a backup.  

Growth Plan

Over the last few years, the startup has received a good traction in the international market, required to convert into potential leads. “We are keeping an eye on choosing the right investment partner, and are focussing on participating in international and national level innovation based competitions with our idea”, he shares.

He adds, “Our strategy has three stages, first and foremost is to focus on bread and butter areas, then we will focus on the product development and third stage is futuristic. That is, with the convergence of next generation AI, blockchain and precision medicine, we would like to move into the fast lane of inorganic growth and become an international leader in AI for healthcare”.

The startup is currently funded from Parthiban’ personal savings, which includes a dedicated facility for operations (capital cost) as well as fixed and variable costs.

What Do You Think?

If you loved this story, do join our Telegram Community.

Also, you can write for us and be one of the 500+ experts who have contributed stories at AIM. Share your nominations here.
What's Your Reaction?
In Love
Not Sure

Copyright Analytics India Magazine Pvt Ltd

Scroll To Top