César de la Fuente
@delafuentelab.bsky.social
Presidential Associate Professor @upenn.edu - using AI to reimagine antibiotic discovery and peptide design. Previously @MIT, @UBC
🔗 https://delafuentelab.seas.upenn.edu
🔗 https://delafuentelab.seas.upenn.edu
Grateful to Elizabeth Stivison for covering our work with superstars Marcelo Torres and Anna Crysler on training AI to uncover novel antibiotics in @asbmb.bsky.social Today! Excited to see growing momentum at the intersection of AI and antibiotic discovery.
November 14, 2025 at 11:13 AM
Grateful to Elizabeth Stivison for covering our work with superstars Marcelo Torres and Anna Crysler on training AI to uncover novel antibiotics in @asbmb.bsky.social Today! Excited to see growing momentum at the intersection of AI and antibiotic discovery.
Thank you Bob Service for highlighting in @science.org our work using AI to uncover antibiotics from archaea. Grateful for the thoughtful coverage—and excited by growing interest in these ancient microbes, a distinct branch of the tree of life.
November 12, 2025 at 8:37 AM
Thank you Bob Service for highlighting in @science.org our work using AI to uncover antibiotics from archaea. Grateful for the thoughtful coverage—and excited by growing interest in these ancient microbes, a distinct branch of the tree of life.
Huge thanks to @nikomccarty.bsky.social and the @asimovpress.bsky.social Podcast for the incredible gift—a book encoded in DNA! 🧬📖
A beautiful fusion of science, art, and storytelling.
A beautiful fusion of science, art, and storytelling.
November 10, 2025 at 9:32 AM
Huge thanks to @nikomccarty.bsky.social and the @asimovpress.bsky.social Podcast for the incredible gift—a book encoded in DNA! 🧬📖
A beautiful fusion of science, art, and storytelling.
A beautiful fusion of science, art, and storytelling.
My friend #EltonBrand @sixersnba.bsky.social and I at my lab's Halloween party talking about the future of AI and Sports. What are your thoughts? Do you think #RudyFernandez and #SergeiIbaka would approve of our costumes?😉 Wishing everyone a Happy Halloween! 🎃
October 31, 2025 at 12:14 PM
My friend #EltonBrand @sixersnba.bsky.social and I at my lab's Halloween party talking about the future of AI and Sports. What are your thoughts? Do you think #RudyFernandez and #SergeiIbaka would approve of our costumes?😉 Wishing everyone a Happy Halloween! 🎃
Big day at @upenn.edu ! Honored to host my friend Elton Brand from @sixersnba.bsky.social for a deep dive into how we’re using AI to fight antimicrobial resistance. Thanks for the sharp questions and team spirit, Elton! 🧪🏀
October 30, 2025 at 2:08 PM
Big day at @upenn.edu ! Honored to host my friend Elton Brand from @sixersnba.bsky.social for a deep dive into how we’re using AI to fight antimicrobial resistance. Thanks for the sharp questions and team spirit, Elton! 🧪🏀
It was an honor to participate in the third course organized by @mobilise.bsky.social @i3suporto.bsky.social, sharing experiences with incredible people. A privilege to learn, share, and connect with such an amazing group.
October 30, 2025 at 10:35 AM
It was an honor to participate in the third course organized by @mobilise.bsky.social @i3suporto.bsky.social, sharing experiences with incredible people. A privilege to learn, share, and connect with such an amazing group.
4/5 As a demo, we used a known antimicrobial peptide as the “key” and designed new candidates: active against multiple pathogens in vitro and reduced A. baumannii burden in mouse infection models—with good tolerability.
October 23, 2025 at 7:50 AM
4/5 As a demo, we used a known antimicrobial peptide as the “key” and designed new candidates: active against multiple pathogens in vitro and reduced A. baumannii burden in mouse infection models—with good tolerability.
3/5 KCM performs across α-helices, β-sheets, mixed, and unstructured regions, delivering precise backbone geometry while staying lightweight and accessible for small labs, startups, and even classrooms.
October 23, 2025 at 7:50 AM
3/5 KCM performs across α-helices, β-sheets, mixed, and unstructured regions, delivering precise backbone geometry while staying lightweight and accessible for small labs, startups, and even classrooms.
2/5 Instead of black-box generation, KCM runs an optimization loop with ESMFold in the mix. You can encode your requirements—geometry, energy, and sequence descriptors—directly in the objective, steering designs in a transparent, controllable way.
October 23, 2025 at 7:50 AM
2/5 Instead of black-box generation, KCM runs an optimization loop with ESMFold in the mix. You can encode your requirements—geometry, energy, and sequence descriptors—directly in the objective, steering designs in a transparent, controllable way.
1/5 Today we are introducing the Key-Cutting Machine (KCM) for protein and peptide design. Like cutting a key to fit a lock, KCM sculpts sequences to match a target backbone—on a single GPU. No retraining. No clusters. 🧬🔑
October 23, 2025 at 7:50 AM
1/5 Today we are introducing the Key-Cutting Machine (KCM) for protein and peptide design. Like cutting a key to fit a lock, KCM sculpts sequences to match a target backbone—on a single GPU. No retraining. No clusters. 🧬🔑
AI could reshape infectious disease control end-to-end—from faster outbreak detection to smarter diagnostics—but progress is constrained by data. My
@thelancetinfdis.bsky.social @thelancet.com
article lays out how to break the bottleneck responsibly.
www.thelancet.com/journals/lan...
@thelancetinfdis.bsky.social @thelancet.com
article lays out how to break the bottleneck responsibly.
www.thelancet.com/journals/lan...
October 15, 2025 at 3:12 PM
AI could reshape infectious disease control end-to-end—from faster outbreak detection to smarter diagnostics—but progress is constrained by data. My
@thelancetinfdis.bsky.social @thelancet.com
article lays out how to break the bottleneck responsibly.
www.thelancet.com/journals/lan...
@thelancetinfdis.bsky.social @thelancet.com
article lays out how to break the bottleneck responsibly.
www.thelancet.com/journals/lan...
(1/2) AI is redefining antibiotic discovery. What once took years with traditional methods can now be done in hours.
September 29, 2025 at 9:24 AM
(1/2) AI is redefining antibiotic discovery. What once took years with traditional methods can now be done in hours.
(6/9)• Multi-objective rewards: predicted MIC + key physicochemical properties
• Rapid & steerable: tune potency and developability on demand
• Amphorium: >2 million machine-annotated candidates for follow-up
• Rapid & steerable: tune potency and developability on demand
• Amphorium: >2 million machine-annotated candidates for follow-up
September 25, 2025 at 11:18 AM
(6/9)• Multi-objective rewards: predicted MIC + key physicochemical properties
• Rapid & steerable: tune potency and developability on demand
• Amphorium: >2 million machine-annotated candidates for follow-up
• Rapid & steerable: tune potency and developability on demand
• Amphorium: >2 million machine-annotated candidates for follow-up
(1/9)For over a century, antibiotics were found by slow, painstaking, serendipitous screening.
Today we share a different path: the first demonstration of deep reinforcement learning (RL) for antibiotic discovery.
Today we share a different path: the first demonstration of deep reinforcement learning (RL) for antibiotic discovery.
September 25, 2025 at 11:18 AM
(1/9)For over a century, antibiotics were found by slow, painstaking, serendipitous screening.
Today we share a different path: the first demonstration of deep reinforcement learning (RL) for antibiotic discovery.
Today we share a different path: the first demonstration of deep reinforcement learning (RL) for antibiotic discovery.
Excited to speak today in beautiful Boulder at the Boulder Peptide Symposium on how AI is accelerating peptide design and discovery.
September 18, 2025 at 10:58 AM
Excited to speak today in beautiful Boulder at the Boulder Peptide Symposium on how AI is accelerating peptide design and discovery.
(4/5)Why it matters: AMR is rising and nature’s dataset is finite. Generative AI lets us explore sequence space evolution never tried—and compress discovery cycles from years to days.
September 2, 2025 at 3:22 PM
(4/5)Why it matters: AMR is rising and nature’s dataset is finite. Generative AI lets us explore sequence space evolution never tried—and compress discovery cycles from years to days.
(3/5)Output: ~50,000 candidates → 46 synthesized → 2 effective against drug-resistant skin infections in mice on par with levofloxacin/polymyxin B, with no detectable adverse effects.
September 2, 2025 at 3:22 PM
(3/5)Output: ~50,000 candidates → 46 synthesized → 2 effective against drug-resistant skin infections in mice on par with levofloxacin/polymyxin B, with no detectable adverse effects.
(2/5)Using diffusion models guided by a protein LM (ESM-2), we sample novel, grammar-consistent sequences and triage them with our APEX AI filters.
September 2, 2025 at 3:22 PM
(2/5)Using diffusion models guided by a protein LM (ESM-2), we sample novel, grammar-consistent sequences and triage them with our APEX AI filters.
Last week, I presented @acs.org in a wonderful session on the future of peptide design to target infectious diseases. With superstar colleagues Terry Moore, Adrian Whitty, Gaurav Bhardwaj, Katerina Leftheris, Jane Aldrich, Elizabeth Rhee, and more!
August 29, 2025 at 9:52 AM
Last week, I presented @acs.org in a wonderful session on the future of peptide design to target infectious diseases. With superstar colleagues Terry Moore, Adrian Whitty, Gaurav Bhardwaj, Katerina Leftheris, Jane Aldrich, Elizabeth Rhee, and more!
Back from Dublin and #IMAP2025—grateful for the chance to share our lab’s work on peptide design and AI. Energized by the science and community. Until next time, Ireland. 🍀
August 28, 2025 at 1:04 PM
Back from Dublin and #IMAP2025—grateful for the chance to share our lab’s work on peptide design and AI. Energized by the science and community. Until next time, Ireland. 🍀
(4/5)These results show that the marriage of deep learning and evolutionary diversity can surface potent therapeutics at digital speed, expanding the antibiotic toolbox when humanity needs it most.
August 12, 2025 at 11:11 AM
(4/5)These results show that the marriage of deep learning and evolutionary diversity can surface potent therapeutics at digital speed, expanding the antibiotic toolbox when humanity needs it most.
(3/5)Of the 80 candidates we synthesized, 93% were active against pathogenic microbes in vitro, and one—archaeasin-73—matched the in-vivo efficacy of last-resort drugs such as polymyxin B.
August 12, 2025 at 11:11 AM
(3/5)Of the 80 candidates we synthesized, 93% were active against pathogenic microbes in vitro, and one—archaeasin-73—matched the in-vivo efficacy of last-resort drugs such as polymyxin B.
(2/5)Our APEX model sifted through hundreds of archaeal proteomes in silico, revealing a previously hidden family of antimicrobial peptides we call “archaeasins.”
August 12, 2025 at 11:11 AM
(2/5)Our APEX model sifted through hundreds of archaeal proteomes in silico, revealing a previously hidden family of antimicrobial peptides we call “archaeasins.”
Inspired by the high school students at @upenn.edu ’s Summer Biomedical Research Academy. Loved talking AI + antibiotics with the next generation of scientists. 🙌🧪🧠
August 4, 2025 at 9:26 AM
Inspired by the high school students at @upenn.edu ’s Summer Biomedical Research Academy. Loved talking AI + antibiotics with the next generation of scientists. 🙌🧪🧠