Listen

Description

What happens when AI begins designing the genetic instructions that control our cells?

In this episode of Galaxy Balance, Dr. Luca Pinello, Associate Professor at Massachusetts General Hospital and Harvard Medical School, joins Cory Smith to explore the intersection of artificial intelligence, genome engineering, and the future of biotechnology.

Luca's lab developed DNA Diffusion, a generative AI system capable of designing regulatory DNA sequences with cell-type-specific activity. These technologies could eventually enable genetic therapies that activate in diseased cells while sparing healthy tissue, offering new possibilities for cancer treatment and precision medicine.

We explore how image-generation AI inspired the development of DNA Diffusion, including an unlikely story involving Italians eating pineapple pizza. Luca also discusses the creation of a new company focused on generative DNA, his team's widely used CRISPResso and CRISPRLungo tools, and the challenge of detecting unintended genome-editing outcomes.

The conversation expands into AlphaGenome, sequence-to-function prediction, AI agents for genomic interpretation, and Luca's vision of conversational genomics, where researchers can ask questions about DNA and receive predictions about its biological function.

We also discuss the biosecurity risks of generative AI, how AI is changing the skills scientists need, and the importance of validating AI-generated discoveries.

Finally, Luca shares how science fiction, particularly Philip K. Dick's Ubik, has influenced his approach to questioning our understanding of reality.

Join us for a conversation about the emerging ability to read, understand, and write the biological programs of life.

 

00:06 Designing DNA elements for specific cellular states

00:36 Meet Dr. Luca Pinello and the Pinello Lab

01:40 From computer science to computational biology

04:33 Biology’s shift toward data, measurement, and interdisciplinary research

06:26 AI safety, transparency, and guardrails

11:02 Biosecurity concerns and dual-use research

15:00 The idea behind DNA Diffusion

19:04 How DNA Diffusion generates and evaluates regulatory DNA

23:11 Genome editing, unintended outcomes, and weighing risks and benefits

29:16 CRISPR Lungo and the value of long-read sequencing

35:16 Building open-source tools to study genome editing

37:55 Academia, industry, and the changing role of the professor

42:41 AI’s effect on mentoring, problem-solving, and verifying research

51:16 Using AI to explore unfamiliar ideas and methods

53:14 Non-coding variants and sequence-to-function models

57:13 Conversational genomics and AI-assisted variant analysis

1:00:37 Science fiction, Ubik, and questioning what we know

1:03:34 Closing thoughts on science and new ways to see the world