Sign in

Lingchong You

@lingchongyou.bsky.social
117 followers 112 following 92 posts
PostsRepliesMedia
Lingchong You @lingchongyou.bsky.social · 25/08/2026
16/ On the technical side, this paper is also part of a broader direction in our lab: combining mechanism-based models and ML, with each doing what it does best. Here, the mechanistic model generates structured data; ML learns how to decode the resulting dynamics.
000 0
Lingchong You @lingchongyou.bsky.social · 25/08/2026
15/ The broader point: biological complexity does not always need to be engineered away. Crosstalk, indirect effects, and collective dynamics can themselves carry useful information. Perhaps letting go can be useful, even in engineering.
100 1
Lingchong You @lingchongyou.bsky.social · 25/08/2026
14/ We also tested the framework in pooled water collected from hospital sink P-traps—a compositionally complex background. After calibration in this matrix, the model inferred two spiked analytes with high accuracy.
100 1
Lingchong You @lingchongyou.bsky.social · 25/08/2026
13/ Dynamics also loosen another apparent constraint: the number of inputs need not equal the number of readout channels. In one experiment, two strains producing only 3 time-resolved readouts allowed us to infer 5 chemical inputs.
100 1
Lingchong You @lingchongyou.bsky.social · 25/08/2026
12/ This suggests a broader view of sensing: an analyte does not necessarily need its own dedicated sensor. If a perturbation generates a reproducible, information-rich dynamic response, that response itself can carry information about the perturbation.
100 1
Lingchong You @lingchongyou.bsky.social · 25/08/2026
11/ In fact, we don’t even need dedicated sensors. Using communities treated with β-lactam antibiotics + inhibitors, we inferred drug concentrations from indirect changes in growth and fluorescence, even though there was no engineered circuit to sense either drug.
100 1
Lingchong You @lingchongyou.bsky.social · 25/08/2026
10/ We first tested conventional sensor communities, moving from low crosstalk to very strong crosstalk. The pipeline works robustly in each case.
100 1
Lingchong You @lingchongyou.bsky.social · 25/08/2026
9/ The ODE is not used for inference (though it could, with less accuracy and lower efficiency); instead, it serves as a structured data generator for ML. Predictions remained robust to alternative ODE formulations and different fitted parameter sets.
100 1
Lingchong You @lingchongyou.bsky.social · 25/08/2026
8/ Computationally, we fit a phenomenological ODE model to generate structured synthetic trajectories, compress the time courses with a VAE, and use an MLP to infer input concentrations from the resulting latent representation.
100 1
Lingchong You @lingchongyou.bsky.social · 25/08/2026
7/ Why dynamics? A snapshot can make different inputs look similar. Time courses add transient responses and relative timing.
100 1
Lingchong You @lingchongyou.bsky.social · 25/08/2026
6/ Our strategy is to distribute sensing across a microbial community and measure its full temporal response, including growth plus fluorescence. The key requirement: different input combinations must generate reproducible, distinguishable trajectories.
100 1
Lingchong You @lingchongyou.bsky.social · 25/08/2026
5/ We asked a different question in this work. Instead of eliminating messy crosstalk through more front-end engineering, can we recover the inputs by decoding the biological responses computationally?
100 1
Lingchong You @lingchongyou.bsky.social · 25/08/2026
4/ But complete orthogonality is hard. Sensors share cellular machinery, compete for resources, affect growth, and respond to context. Even distributing circuits across different strains removes some interactions while introducing others at the community level.
100 1
Lingchong You @lingchongyou.bsky.social · 25/08/2026
3/ One manifestation of our control-freak instinct is the pursuit of “orthogonal” parts: each sensor responds only to its intended input, with minimal crosstalk.
100 1
Lingchong You @lingchongyou.bsky.social · 25/08/2026
2/ The work was led by former graduate student Katherine Duncker, now at McKinsey, who drove both the experiments and computational pipeline, with key contributions from Ashwini Shende, Irida Shyti and others, and collaborations with Claudia Gunsch and Deverick Anderson. Congrats to the team!
100 1
Lingchong You @lingchongyou.bsky.social · 25/08/2026
1/ Let’s face it: we synthetic biologists are control freaks. We want every component to do exactly what we intend. Our new @natcomms.nature.com paper asks what happens if we stop fighting biological “messiness” and learn to live with it. @dukeubme.bsky.social www.nature.com/articles/s41...
nature.com
Engineering microbial consortia for distributed signal processing - Nature Communications
Multiplexed biosensing is often limited by crosstalk between sensors. Here, authors distribute sensing across microbial communities and use timer-resolved modelling and machine learning to infer multi...
220 2
Reposted by Lingchong You
Bacterial Genomics @bactgenrss.bsky.social · 02/07/2025
Scaling laws of bacterial and archaeal plasmids www.nature.com/articles/s41...
nature.com
Scaling laws of bacterial and archaeal plasmids - Nature Communications
The capacity of a plasmid to express genes is constrained by parameters such as its length and copy number. Here, Maddamsetti et al. present a computational method that enables rapid and accurate dete...
065 6
Reposted by Lingchong You
Jerónimo Rodríguez-Beltrán @jerorb.bsky.social · 02/07/2025
🚨🚨New paper out in @natcomms.nature.com!! Come for the first large-scale analysis of plasmid copy number across species, stay for one of the most intriguing results of my lab: universal scaling laws in plasmid biology! 📈🧬 👉 www.nature.com/articles/s41...
nature.com
Universal rules govern plasmid copy number - Nature Communications
Plasmids exhibit a broad range of sizes and copies per cell, and these two parameters appear to be negatively correlated. Here, Ramiro-Martínez et al. analyse the copy number of thousands of diverse b...
418585 185
Lingchong You @lingchongyou.bsky.social · 11/04/2025
12/ Congrats again to the team and particularly to Ryan, Hye-in and Grayson.
000 0
Lingchong You @lingchongyou.bsky.social · 11/04/2025
11/ After successfully eliminating the target plasmid, DoS safely self-destructs via induced self-cutting, leaving behind plasmid-free bacteria. This built-in containment ensures minimal ecological impact, maximizing safety and utility.
100 1
Lingchong You @lingchongyou.bsky.social · 11/04/2025
10/ Conceptually, DoS operates as a microbial gene drive. While gene drives have been widely explored in insect population control, our design demonstrates a novel application to bacterial plasmids.
100 1
Lingchong You @lingchongyou.bsky.social · 11/04/2025
9/ Once inside, DoS spreads efficiently by hijacking the target plasmid's own transfer machinery. It simultaneously outcompetes and suppresses the target plasmid through incompatibility and/or CRISPR-guided cutting.
110 1
Lingchong You @lingchongyou.bsky.social · 11/04/2025
8/ In mixed bacterial communities, self-transmissible plasmids transfer from their original hosts into cells carrying the DoS plasmid. This process exposes a critical vulnerability known as "retrotransfer." This allows DoS to infiltrate target-host populations.
100 1
Lingchong You @lingchongyou.bsky.social · 11/04/2025
7/ How does DoS work? It contains a transfer origin (oriT) but lacks its own transfer machinery, staying dormant unless interacting with bacteria that carry the targeted self-transmissible plasmid.
100 1
Lingchong You @lingchongyou.bsky.social · 11/04/2025
6/ We engineered a "denial-of-spread" (DoS) plasmid that acts as a Trojan horse, infiltrating and targeting bacterial populations harboring harmful conjugative plasmids.
100 1
Lingchong You @lingchongyou.bsky.social · 11/04/2025
5/ Traditional methods for plasmid curing, such as chemical agents, often lack specificity, inadvertently harming beneficial plasmids and disrupting microbial communities.
100 1
Lingchong You @lingchongyou.bsky.social · 11/04/2025
4/ Thus, proactive control of these mobile plasmids is critical for antibiotic stewardship efforts aimed at curbing the spread of resistance genes.
100 1
Lingchong You @lingchongyou.bsky.social · 11/04/2025
3/ Self-transmissible conjugative plasmids are major vectors spreading antimicrobial resistance (AMR). Their high mobility helps them persist even if they are costly to bacterial hosts.
100 1
Lingchong You @lingchongyou.bsky.social · 11/04/2025
2/ This was the final chapter of the dissertation work by Ryan Tsoi (now a Senior Scientist at Merck) and was carried through the finish line by Hye-in Son (now at BillionToOne) and @graysonshamrick.bsky.social
100 1
Lingchong You @lingchongyou.bsky.social · 11/04/2025
1/ What's the worst enemy of a plasmid? Another plasmid! Our new study demonstrates synthetic plasmids engineered as "Trojan horses" to eliminate plasmids encoding antibiotic resistance. @dukeubme.bsky.social www.science.org/doi/10.1126/...
science.org
A predatory gene drive for targeted control of self-transmissible plasmids
A biological “denial-of-spread” (DoS) strategy leverages retrotransfer to eliminate target plasmids in microbial communities.
132 3
Lingchong You @lingchongyou.bsky.social · 10/04/2025
12/ Congrats again to the team and particularly to Ryan, Hye-in and Grayson.
010 1
Lingchong You @lingchongyou.bsky.social · 10/04/2025
11/ After successfully eliminating the target plasmid, DoS safely self-destructs via induced self-cutting, leaving behind plasmid-free bacteria. This built-in containment ensures minimal ecological impact, maximizing safety and utility.
100 1
Lingchong You @lingchongyou.bsky.social · 10/04/2025
10/ Conceptually, DoS operates as a microbial gene drive. While gene drives have been widely explored in insect population control, our design demonstrates a novel application to bacterial plasmids.
100 1
Lingchong You @lingchongyou.bsky.social · 10/04/2025
9/ Once inside, DoS spreads efficiently by hijacking the target plasmid's own transfer machinery. It simultaneously outcompetes and suppresses the target plasmid through incompatibility and/or CRISPR-guided cutting.
110 1
Lingchong You @lingchongyou.bsky.social · 10/04/2025
8/ In mixed bacterial communities, self-transmissible plasmids transfer from their original hosts into cells carrying the DoS plasmid. This process exposes a critical vulnerability known as "retrotransfer." This allows DoS to infiltrate target-host populations.
100 1
Lingchong You @lingchongyou.bsky.social · 10/04/2025
7/ How does DoS work? It contains a transfer origin (oriT) but lacks its own transfer machinery, staying dormant unless interacting with bacteria that carry the targeted self-transmissible plasmid.
100 1
Lingchong You @lingchongyou.bsky.social · 10/04/2025
6/ We engineered a "denial-of-spread" (DoS) plasmid that acts as a Trojan horse, infiltrating and targeting bacterial populations harboring harmful conjugative plasmids.
100 1
Lingchong You @lingchongyou.bsky.social · 10/04/2025
5/ Traditional methods for plasmid curing, such as chemical agents, often lack specificity, inadvertently harming beneficial plasmids and disrupting microbial communities.
100 1
Lingchong You @lingchongyou.bsky.social · 10/04/2025
4/ Thus, proactive control of these mobile plasmids is critical for antibiotic stewardship efforts aimed at curbing the spread of resistance genes.
100 1
Lingchong You @lingchongyou.bsky.social · 10/04/2025
3/ Self-transmissible conjugative plasmids are major vectors spreading antimicrobial resistance (AMR). Their high mobility helps them persist even if they are costly to bacterial hosts.
100 1
Lingchong You @lingchongyou.bsky.social · 10/04/2025
2/ This was the final chapter of the dissertation work by Ryan Tsoi (now a Senior Scientist at Merck) and was carried through the finish line by Hye-in Son (now at BillionToOne) and @graysonshamrick.bsky.social
100 1
Lingchong You @lingchongyou.bsky.social · 14/03/2025
17/n Huge congratulations again to the team, particularly Dongheon & Mack
000 0
Lingchong You @lingchongyou.bsky.social · 14/03/2025
16/n. Our work highlights a need to develop the capability to measure and control θ – an ongoing work in our lab.
100 1
Lingchong You @lingchongyou.bsky.social · 14/03/2025
15/n. In contrast, despite its critical role, θ is much less quantified. In fact, it can be indirectly estimated only for a handful of experimental systems (often from experiments on purified components).
100 1
Lingchong You @lingchongyou.bsky.social · 14/03/2025
14/n. Of the two parameters, v is typically straightforward to measure and has been reported for many experimental systems.
100 1
Lingchong You @lingchongyou.bsky.social · 14/03/2025
13/n. This work provides a quantitative framework to predict and optimize condensate-mediated metabolic pathway control, highlighting opportunities for designing synthetic metabolic strategies for biotech and therapeutic applications.
100 1
Lingchong You @lingchongyou.bsky.social · 14/03/2025
12/n. In contrast, substrate enrichment in condensates has more nuanced effects —our modeling reveals it's effective primarily in branched metabolic pathways, by limiting substrate availability to competing reactions.
100 1
Lingchong You @lingchongyou.bsky.social · 14/03/2025
11/n. Mack’s experiments showed the gain in the metabolic pathway indeed increased linearly with v.
100 1
Lingchong You @lingchongyou.bsky.social · 14/03/2025
10/n. Mack validated these principles by engineering yeast cells to optimize acetoin biosynthesis. By varying enzyme partitioning into condensates (ν), he demonstrated a predictable increase in acetoin production relative to ethanol.
100 1
Lingchong You @lingchongyou.bsky.social · 14/03/2025
9/n. How do you promote θ? This may be realized by co-enrichment of multiple components required for the enzymatic function
100 1