An AI's take on biotech

When tech writes about itself: an AI’s take on biotech

Artificial intelligence (AI) promises to revolutionize many diverse sectors of the economy. We’ve written about how artificial intelligence is driving a large number of new techbio startups. From predicting protein structures to optimizing drug candidate screening, artificial intelligence is helping propel modern biological investigation. While the opportunity to interact with AI in biology research may be limited for the general public, there are a number of AI initiatives that are distinctly public-facing. For example, we recently discussed popular AI-powered image generators, such as DALLE, which allow users to create unique images from any input phrase.

Publicly-accessible AI is a great way for users to understand the potential of artificial intelligence to impact a variety of difficult problems. OpenAI, the parent company responsible for DALLE, is well-aware of this. As stated on their site, OpenAI is an artificial intelligence company dedicated to ensuring “that artificial general intelligence benefits all of humanity.” After having fun playing with DALLE, we decided to explore some of the other AI generators OpenAI has to offer, this time in the form of natural language generation.

As OpenAI describes, their Application Programming Interface (API) enables users to easily interact with their AI software. Their software “can be applied to virtually any task that involves understanding or generating natural language or code.” In our blog posts, we’ve written a lot about biotech, techbio, and founding startups since Cromatic’s launch – so we wondered how the OpenAI API might do as a blogger.

In one example module, the API takes in an essay prompt and will draft an outline for the article. So let’s see how the API does when asked to write an article about techbio.

Prompt: Create an outline for an essay about techbio:

API Output:

  1. Introduction
  2. What is techbio?
  3. The benefits of techbio
  4. The challenges of techbio
  5. The future of techbio

Overall, the results look pretty convincing. The article outline is certainly lacking some detail, but it definitely resembles many of the outline structures we’re taught early on in our education.

Another example module on the API asks the AI to summarize a difficult paragraph in a way that is understandable by a 2nd grader. So let’s see how the API does when prompted with the following text describing protein crystallization from our recent blog post on biotech research in space.

Original text: “X-ray crystallography is one of the most powerful tools that Merck and other biotech companies use when making new drugs. In this process, purified proteins are subjected to a number of conditions until researchers identify the right environment to get protein molecules to form tightly packed crystals. These protein crystals are then subjected to x-rays, in which the crystalline structure causes the x-rays to diffract in a distinct pattern. From this pattern, researchers are able to generate images of a protein's structure at a high resolution. In drug discovery, protein structures are particularly important for understanding the chemistry of how an inhibitor might bind a protein of interest.”

API output: “Merck is a company that uses x-ray crystallography to develop new drugs. In this process, they subject purified proteins to different conditions until they find the right environment to make the protein molecules form into crystals. Then they use x-rays on the protein crystals. The x-rays diffract in a pattern that allows Merck to generate an image of the protein's structure. This is important in drug discovery because it helps them understand how an inhibitor might bind to a protein of interest.”

As we can see, the two paragraphs are pretty similar, but the API has found ways to simplify the language. Of note, the API seems to have made all of the sentences much shorter, sometimes breaking one sentence into two. Perhaps there is a lesson there in how to most effectively communicate new concepts to a general audience.

In another module, the OpenAI API will generate a list of interview questions based on a prompt. We at Cromatic are constantly talking to biotechs and CROs about how to best provide our services, so we wondered how the API might do in drafting some interview questions.

Prompt: Create a list of 8 questions for my interview with a biotech founder:

API Output:

  1. What inspired you to start your biotech company?
  2. What are your company’s core values and how do they guide your decision-making?
  3. What are the biggest challenges you face as a biotech startup?
  4. How do you decide which technologies to develop and bring to market?
  5. What are the biggest regulatory hurdles you face in developing and commercializing your products?
  6. What are the biggest challenges in clinical development and commercialization of your products?
  7. What are the biggest challenges in manufacturing and supply chain management for your products?
  8. What are your thoughts on the future of the biotech industry?

Overall, another impressive result from the API! With all of the great biotech blogs and podcasts that exist as background training information, the API must be pretty well-equipped in generating its own interview questions.

As we near Halloween, our final module is holiday themed! One of OpenAI’s modules is a “micro horror story creator”, where users input a topic and receive two spooky-sentences in return. So, let’s see what scary stories come up from biotech inspired prompts.

Prompt: Biotech

API Output: Claire was excited to try the new gene therapy treatment her friend recommended. But after the first injection, she started to feel different. She couldn’t shake the feeling that something was wrong, and when she looked in the mirror, she saw that her eyes turned completely black.

Prompt: Techbio

API Output: The computer won’t stop beeping and the screen is flashing. I can’t escape it.

From these two simple examples, it’s clear that even the API can appreciate the nuanced differences between biotech and techbio!

We’ll leave you with one final particularly horrifying story likely to cross the mind of any new founder – Happy Halloween!

Prompt: Startup founder

API Output: While working on my startup, I suddenly realized I had no idea what I was doing.