First of all, thank you for subscribing to the newsletter 🙏!

AI is definitely on fire 🔥. So many exciting things are happening in the realm of “AI + [field]”. I cannot wait to share them with you.

💉 AI + mRNA

Today (October 2, 2023), Katalin Karikó and Drew Weissman won the Nobel Prize “for their discoveries concerning nucleoside base modifications that enabled the development of effective mRNA vaccines against COVID-19.” There is so much AI can contribute to mRNA and RNA-based therapeutics!

Just a few days ago, on September 27, 2023, Deep Genomics announced their RNA large language model (LLM), called BigRNA. The training process combines convolution and attention to predict RNA-Seq coverage and junctions, which is quite different from LLM training in natural language procession (NLP) in terms of algorithm and prediction task. It might be superior in predicting the effects of variants on gene expression, but I’m keen to see how it compares with RNA LLMs trained using traditional methods, like, like RNABERT. The manuscript was published on bioRxiv.

Earlier in September 2023, an AI-powered startup, Inceptive, raised $100 million from NVIDIA and Andreessen Horowitz. Founded by Jakob Uszkoreit, one of the inventors of the transformer model, Inceptive is using AI to design mRNA-based “biological software”.

If you are interested in AI + RNA, Kaggle is hosting a competition on RNA structure prediction, from September 7 to December 7, 2023. It is an excellent opportunity to showcase, practice, and learn your skills. You have two months to go. (alert: RNA is quite different from protein. In the 2022 CASP15 competition, top RNA structure prediction methods did not use deep learning.)

We all know the mRNA story from Moderna and BioNTech. Many more RNA-based therapeutics could significantly benefit from RNA LLM and structure prediction, including mRNA, circular RNA, antisense oligonucleotides (ASO), RNA interference (RNAi), RNA editing, RNA splicing- or structure-targeted small molecule, etc.

👩🏽‍⚕️ AI + Healthcare

ChatGPT can now see, hear, and speak?! We need multimodal capability in healthcare too! There are more data modalities in healthcare, including omics, clinical notes, medical images, and wearable sensors, that capture the complexity of human health and disease. In a recent article in Science, Eric Topol highlighted the applications and needs of multimodal AI in medical fields. He also published on the same topic in Nature Medicine last year with more details (Figure 1).

Google may already have a good solution, called Med-PaLM Multimodal (Med-PaLM M). Basically, text and tabular data are covered by the language transformer and all other data are treated as an image and fed to a vision transformer. The manuscript was published on Arxiv, but unfortunately, they will not open-source the code and model weights. Since it is quite an important work, we will dive more into the technical details in a later newsletter.

📊 AI + Statistics

Do statisticians hate the hype of AI? A recent NEJM review article, titled “Where Medical Statistics Meets Artificial Intelligence”, commented that statistics and AI are very different. Yet, both are useful and powerful in medical studies. The authors suggested that “statisticians should embrace AI, and in response, the field of AI will benefit from increased statistical thinking.” As someone trained in both fields, I cannot agree more and highly recommend everyone reads the entire article.

🧫 AI + Science

Last week, Nature launched a new series about AI + Science, named “Science and the New Age of AI“. In the series, you can learn:

• What other scientists think about AI (spoiler alert: it’s from a survey result so the data might be biased);

• How to fight against deepfakes generated by AI;

• How to leverage the right data and AI to speed up drug discovery (Figure 2);

• and more, as Nature will continuously publish more “AI + Science” editorial articles in the coming weeks and months (We will keep an eye on it.).

If you find the newsletter useful, please consider:

• 🔊 Sharing the newsletter with other people

• 👍 Upvoting on Product Hunt

• 📧 Sending any feedback, suggestions, and questions by directly replying to this email or writing reviews on Product Hunt

• 🙏 Supporting us with a cup of coffee.

Thanks, and see you next time!