SC
Arc Institute's Patrick Hsu on Building an App Store for Biology with AI
@ Sequoia Capital
04/15/25
Takeaways
PH
Patrick Hsu
04/15/25
@ Sequoia Capital
The model can interpret genetic mutations and provide insights into their functional consequences, which is state-of-the-art in computational biology.
PH
Patrick Hsu
04/15/25
@ Sequoia Capital
EVO2 is a revolutionary biological foundation model that interprets and generates genomic sequences, identifying patterns in genetic code and predicting the effects of mutations that can impact health.
PH
Patrick Hsu
04/15/25
@ Sequoia Capital
My vision extends beyond just creating better drugs; I aim to build a comprehensive understanding of biology at all scales.
PH
Patrick Hsu
04/15/25
@ Sequoia Capital
Machine learning for biology is not limited to drug design; it encompasses a broader potential that significantly impacts human lives.
PH
Patrick Hsu
04/15/25
@ Sequoia Capital
We have a unifying theory for biology, which is evolution, that influences all levels of biological organization, from ecosystems to individual molecules.
PH
Patrick Hsu
04/15/25
@ Sequoia Capital
We are training models inspired by evolutionary forces to connect biological sequences directly to their functions, leveraging the effects of natural selection through DNA mutations.
PH
Patrick Hsu
04/15/25
@ Sequoia Capital
We started with DNA because it is the fundamental information layer of life and where we have the most data available from the scientific community.
PH
Patrick Hsu
04/15/25
@ Sequoia Capital
Our model, EVO, is designed to predict the next base in a DNA sequence, which can determine health outcomes, such as the presence of sickle cell anemia mutations.
PH
Patrick Hsu
04/15/25
@ Sequoia Capital
The model can also predict the next amino acid residue, which is crucial for determining the functionality of enzymes in biological processes.
PH
Patrick Hsu
04/15/25
@ Sequoia Capital
The model learns from vast datasets of genomic sequences, allowing it to identify biologically meaningful patterns that can lead to new insights in biology.
PH
Patrick Hsu
04/15/25
@ Sequoia Capital
The tools we develop become useful when they lower the barrier to adoption, similar to how tools like BLAST and AlphaFold are widely used in biology.
PH
Patrick Hsu
04/15/25
@ Sequoia Capital
ARC aims to foster long-term research capabilities and breakthroughs in biology by leveraging multidisciplinary expertise and long-term thinking.
PH
Patrick Hsu
04/15/25
@ Sequoia Capital
Thirty years ago, the interests of basic science and industry were quite different, but today they heavily overlap, with both fields focusing on similar questions and challenges.
PH
Patrick Hsu
04/15/25
@ Sequoia Capital
In academia, there is less focus on hardcore drug manufacturing aspects like pharmacokinetics or toxicology, but there is significant interest in molecular glues, machine learning models, and new drug concepts.
PH
Patrick Hsu
04/15/25
@ Sequoia Capital
The differences between academia and industry are often overstated; both operate within overlapping distributions today, although they have fundamentally different incentives that drive their behaviors.
PH
Patrick Hsu
04/15/25
@ Sequoia Capital
At Arc Institute, we blend our academic side with a technical staff structured more like industry, aiming to create a model that others might want to replicate.
PH
Patrick Hsu
04/15/25
@ Sequoia Capital
We aim to create technologies and platform capabilities that are broadly useful, enabling new mechanistic insights and potentially curing diseases.
PH
Patrick Hsu
04/15/25
@ Sequoia Capital
AI can enhance efficiency in various steps of drug development, such as target identification, data analysis, and literature review, ultimately aiming to improve the overall success rate of drug discovery.
PH
Patrick Hsu
04/15/25
@ Sequoia Capital
The pharmaceutical industry has a low probability of success, around 10%, which drives the need for AI to improve efficiency in drug discovery and development processes.
PH
Patrick Hsu
04/15/25
@ Sequoia Capital
The promise of AI in drug discovery lies not just in predictive power but in the ability to simulate experiments in a parallelized fashion, which could significantly reduce the time and cost involved in traditional methods.
PH
Patrick Hsu
04/15/25
@ Sequoia Capital
The best scientists are those who can generate creative ideas and execute them, often by making connections between disparate fields that others might overlook. To achieve this, you need to read a lot of papers across various domains, including chemical biology, molecular biology, and AI, to traverse the boundaries between disciplines.
PH
Patrick Hsu
04/15/25
@ Sequoia Capital
Building a multidisciplinary team is challenging because there are more people in machine learning and biology than those who are truly bilingual in both fields.
PH
Patrick Hsu
04/15/25
@ Sequoia Capital
At Arc Institute, I look for individuals who think about science outside of the lab, have completed end-to-end projects, and possess the grit to see things through to completion.
PH
Patrick Hsu
04/15/25
@ Sequoia Capital
The last 20% of a project often requires 80% of the work, emphasizing the importance of finishing and refining projects multiple times.
PH
Patrick Hsu
04/15/25
@ Sequoia Capital
At Arc Institute, we are focused on applying high-throughput scalable technologies to understand multi-system interactions, particularly at the neuro-immune interface.
PH
Patrick Hsu
04/15/25
@ Sequoia Capital
Interoception, the sense of internal body states, is crucial for understanding how our bodies communicate with our brains, influencing our feelings and behaviors.
PH
Patrick Hsu
04/15/25
@ Sequoia Capital
Programming biology goes beyond drug development; it involves understanding how hormones can influence thoughts, feelings, and behaviors in powerful ways.
PH
Patrick Hsu
04/15/25
@ Sequoia Capital
Exercise can trigger a specific gut bacteria to release molecules that signal the brain to release dopamine, creating a functional circuit for runner's high.
PH
Patrick Hsu
04/15/25
@ Sequoia Capital
Deep psychological stress can lead to gut inflammation through signaling from the brain, illustrating the brain-body connection and its impact on health.
PH
Patrick Hsu
04/15/25
@ Sequoia Capital
The future of drug development will involve a holistic approach, integrating functional medicine with traditional drug therapies to enhance overall health.
PH
Patrick Hsu
04/15/25
@ Sequoia Capital
In the future, we will have comprehensive health panels that allow individuals to make informed decisions about their health based on integrated data.
PH
Patrick Hsu
04/15/25
@ Sequoia Capital
AI doctors will be able to integrate multimodal information, providing personalized health recommendations based on real-time data.
PH
Patrick Hsu
04/15/25
@ Sequoia Capital
The integration of genetic information with health biomarkers will lead to more accurate predictions of health outcomes, moving beyond traditional methods.
PH
Patrick Hsu
04/15/25
@ Sequoia Capital
We need to connect genetic signatures with health data to make meaningful predictions about individual health and lifestyle choices.
PH
Patrick Hsu
04/15/25
@ Sequoia Capital
The future of personalized health will involve a system that connects genetic data with lifestyle factors to provide tailored health recommendations.
PH
Patrick Hsu
04/15/25
@ Sequoia Capital
The goal is to create a system that can analyze vast amounts of health data and provide personalized insights for better health management.
PH
Patrick Hsu
04/15/25
@ Sequoia Capital
By 2025, we will be able to design full antibodies and enzymes through advanced AI models, revolutionizing drug development.
PH
Patrick Hsu
04/15/25
@ Sequoia Capital
In the next few years, we aim to build a comprehensive database of virtual cells to enhance drug target design and therapeutic success rates.
PH
Patrick Hsu
04/15/25
@ Sequoia Capital
By 2050, we envision a fully automated lab capable of simulating drug interactions with high accuracy, significantly speeding up drug discovery processes.
PH
Patrick Hsu
04/15/25
@ Sequoia Capital
We need to conduct more human experiments to improve our understanding of drug effects and their mechanisms in the human body.
PH
Patrick Hsu
04/15/25
@ Sequoia Capital
Despite the challenges in research, it's important to have fun and maintain a positive outlook to foster persistence and achieve long-term goals.