Generative AI & Biotech's Future - Interview with Startups of the Year Nominee, NExtNet Inc.

Hey Hackers,

NExTNet Inc. has been nominated in HackerNoon's annual Startup of the Year awards in San Francisco, USA.

Please vote for us here.

Read more about us below to understand why we deserve your vote.

Meet NExTNet Inc.

NExTNet is a Venture-backed Silicon Valley-based Enterprise software startup developing Generative AI technologies for mission-critical applications in biotech and pharma. Our mission is to organize and integrate the world's biomedical knowledge and make it accessible. Imagine NExTNet = Oracle x Palantir. Meet our team here.

My Role

In late 2017, I founded Mekonos, a fast-growing biotechnology platform company developing ground-breaking silicon chip technologies to accelerate the development of personalized medicine. Imagine Nvidia for biotech. I founded NExTNet in late 2020 frustrated by the extremely high technical barrier to accessing the scattered knowledge buried within mountains of multi-modal and disparate biomedical data and information. Our goal is to democratize access to the world's biomedical knowledge to accelerate drug discovery and development. There are more than 23,500 diseases known to mankind, of which ~3% have some form of cure or treatment.

How We're Disrupting the ‘Intersection of AI and Biomedical Industry

Roughly 95% of the world’s data have been generated in the last 5-10 years. The emergence of high-resolution, multi-modal biomedical data (10s of millions of scientific publications, patents, grants, sequencing data, gene and protein expression, drug compounds, biochemical pathways, diseases, imaging, etc.) at scale has resulted in an enormous amount of human knowledge scattered across data silos. Querying all that knowledge has become remarkably complex.

At NExTNet, we use our proprietary Large Language Models stack to mine for associations between scientific content in text (e.g., published literature, clinical trial record, patents) and other public and proprietary databases (gene sequencing, protein expression, diseases, pathways, pathogens, drugs, imaging...) and semantically link them into the world's fastest-growing scientific semantic web. The technical barrier to query and access such scattered knowledge is scarily high: an average researcher would need to have extensive knowledge of the command-line interface by mastering myriads of R libraries, learning a whole suite of Python packages or architecting complex queries using languages such as SQL, SPARQL, etc.

NExTNet's cloud-native platform allows these researchers and scientists to ask and answer complex questions in simple natural language without having to master coding, querying languages, or arcane statistics. Our commercial and technological breakthrough is enabling scientists to identify patterns hidden deep within disparate and multi-modal datasets, ranging from scientific papers and clinical trial reports to gene sequencing and protein expression atlases -- so that they can quickly find leads buried in mountains of information via our Graphical User Interface (GUI). We are disrupting the intersection of Language and Generative Artificial Intelligence and the Biomedical industry.

Standing Out from The Crowd

Knowledge discovery for life sciences R&D is an embryonically developing market. The key problem in biomedicine today:

• Biomedical knowledge is growing at an exponential rate.

• As the amount of disparate data (e.g., scientific text, molecular data from sequencing, gene editing and other experiments etc.) increases, it becomes harder and harder for researchers to stay up-to-date with the state of the art.

• Since the technical barrier to access knowledge buried deep within all these multi-modal data silos is very challenging without the knowledge of the command-line interface, decisions are made without key insights and missed knowledge.

  
  

At NExTNet, our 10x value increase over competitors is our state-of-the-art Generative AI pipeline not only reads 10s of millions of scientific text (full-length peer-reviewed publications, abstracts, patents, grants etc.) but also analyzes a massive corpus of molecular databases (genes, proteins, and pathways) and mines for hidden associations and patterns across these disparate data (public domain and proprietary). Our competitors are just limited to text mining without expanding into other modalities of data.

Our Predictions/Thoughts on the Drug Discovery and Development/ Biomedical Research Industry in 2023

• As a startup, we are laser-focusing on the drug discovery and development market due to the heavy $$$ being spent before any clinical trial is done in the pre-clinical phases over the course of 3-7 years at least. There is also a significant amount of VC money being invested in AI drug discovery companies (Atomwise, Insilico, Benevolent AI, Exscientia, Insitro, Recursion, etc. that have their own therapeutic pipeline).

  
  

• Given the significant risks associated with bringing a drug to market coupled with long development timelines (that can stretch up to a decade or even more), engaging with regulatory agencies such as the FDA, etc., and long cycle time of co-development partnerships (e.g. upfronts + milestones + royalties), we decided early on to just focus on the software development and rather enable other biotech and pharma companies developing cutting-edge therapeutics 'contextualize' their proprietary data on top of NExTNet's semantic knowledge network --> ask and answer complex biological questions to advance their drug development efforts. A company such as Benchling has found significant success so far as an Enterprise software company.

  
  

• To give a sense of the massive cost of drug discovery and development: 30 years ago, it took ~$300M USD to bring a drug to market v/s today it takes ~$2.4B USD (accounting for inflation). This is not because any single stage of the discovery and development process for a given molecule has gotten that much more expensive. In fact, there have been meaningful increases in productivity for a number of these stages -- so if anything the total cost of bringing a drug to market should have gotten cheaper. The real reason for such high costs in bringing a drug to market is that most of the drugs that we put in every single phase fail and the funnel has gotten progressively wider and wider narrowing down to a smaller and smaller number of drugs that make it to the next phase.

  
  

• This is clearly a failure of our ability to look at say characteristics (e.g., understanding molecular mechanisms of action, biological processes, etc.) at one stage and predict what's going to succeed in the next stage. This is where better predictive ML models (e.g. natural language AI) can help tremendously.

  
  

• In addition, diseases have traditionally been diagnosed by their symptoms and their location in the body and not by the underlying molecular mechanisms, pathways, or biological processes unique to a specific patient. Identifying targets and discovering connections with pathways early on in the disease progression helps with intervening at the causal stage rather than the symptoms. Hence a lot of the prescribed medicines don't often work on patients.

What word defines the state of the ‘intersection of AI and Biomedical Industry’ in 2023?

Language AI

Why we decided to participate in HackerNoon's Startup of the Year awards

We got invited! :)

Final Thoughts

Biomedical knowledge is immense and growing quickly. The publicly available knowledge is disconnected - it is impossible to identify patterns/associations hidden deep within such disconnected knowledge. A company's internal knowledge is disconnected from it too. This presents a significant challenge for scientists in the biomedical industry because they:

• Miss important and potentially relevant knowledge

• Unable to stay up to date or keep up with new knowledge

• Decisions are made without the advantage of complete insights

  
  

At NExTNet, we are building the best-in-class Generative/Language AI stack to enable scientists to identify patterns hidden deep within disparate and multi-modal datasets, ranging from scientific papers and clinical trial reports to gene sequencing and protein expression atlases -- so that they can quickly find leads buried in mountains of information.

Start your biomedical knowledge discovery journey by signing up for free right here.