Devansh Pandey

Half of people with heart disease first find out after a heart attack. What if we could see risk years earlier using MRI, genetics, and blood tests? We introduce preCog, a multimodal AI for preventive cardiology trained on >500k images and genetic data from 1.25M individuals.

🧵Super excited to share our new paper in @NatureGenet ⭐️ We apply a new multiple traits and ancestry approach to improve genetic discovery and polygenic prediction in lung diseases and traits 🫁

@he_yix Amazing work @he_yix

Congratulations to @EucharistKun for successfully defending his PhD thesis on applying deep learning methods to >1.2M images from 60k people and connecting this to genetic data to understand aging and development. He's on the postdoc job market: scholar.google.com/citations?user….

I will be presenting a poster about the LGM in Europe and the Mesolithic across Eurasia from 2:30 to 4:30 this Friday at ASHG (poster # 6027). It is unpublished and thus subject to change! If you're interested in chatting, please find me there, reach out here, or email me!

Big thanks to my amazing co-authors for their contributions and collaboration on this project. @xliaoyi, @EucharistKun, @joyce_yiyi_wang, @ChenfeiLi2000, @melek_alaa, Julie C. Dicarlo, Edward Castillo, @LubbDup, Charles A. Taylor and @vagheesh

I will be presenting this work at #ASHG25 in: 🧬 Session 77 — Beyond Sequencing: Genetic Architecture of Complex Traits 📅 Friday, Oct 17 | 1:30–2:30 PM ET and also as a poster presentation: 📍 Poster Board: 9080F 🗓 Friday, Oct 17 | 2:30–4:30 PM ET

Half of people with heart disease first find out after a heart attack. What if we could see risk years earlier using MRI, genetics, and blood tests? We introduce preCog, a multimodal AI for preventive cardiology trained on >500k images and genetic data from 1.25M individuals.

New paper on our group applying deep learning to measure spine curvature in thousands of individuals to identify new genetic loci associated with susceptibility. Led by star undergrad Michael Zeosky who is applying to PhD programs this fall! x.com/EucharistKun/s…

Scoliosis is the most common spinal deformity, but its genetic causes are still unclear. Using AI to measure spinal curves in 57,887 people, we discovered 2 novel genetic risk loci - showing how deep learning can unlock insights into musculoskeletal health medrxiv.org/content/10.110…

Most genetic reports label variants as ✅ “pathogenic” or ❌ “benign.”
But what patients really want to know is: what are my chances of getting the disease?
🚨 In our new Science study, we used AI + 1.3M patient records to better answer that…

What did family trees look like in prehistory? Ancient DNA now lets us sample hundreds of people from a site, but pedigrees have been difficult to reconstruct. Our new tool “repare” finally enables automatic ancient pedigree reconstruction… more in the replies! ⬇️