Biology subjects

Esvelt, K.

Publications and source records attributed to Esvelt, K..

8 recordsLinked to original sources

bioRxiv2026

Developing a Standard Definition for Sequences of Concern

Readily available nucleic acid synthesis is both critical for the bioeconomy and an increasingly pressing security concern due to the potential for accidental or deliberate misuse. While biosecurity experts broadly agree that nucleic acid providers should screen orders for potential "sequences of concern," there has previously been no agreed standard for how to define and recognize such sequences. To address this gap, we first organized a test set of 1.1 million sequences from pathogens and toxins on the Australia Group Common Control Lists and their non-controlled relatives, along with model organisms and synthetic constructs. An initial categorization of sequences as to whether or not they were sequence of concern was produced by comparing the results of four biosecurity screening systems for each of these sequences, finding that these systems already agreed on the categorization of more than 80% of sequences. We then refined these results through a science-based stakeholder review process to define a rubric for determining whether a sequence should be flagged as a potential sequence of concern, then applied this rubric to improve the categorization of test sets. The result is a rubric that identifies sequences of concern with respect to human pandemic-potential viruses, key classes of low-risk genes, and controlled toxins. Applying this rubric to the test set collection has reduced the number of test sequences with disputed categorization by 44.3% for controlled viruses and 10.7% across the test set as a whole. Together, these results provide a concrete "sequence of concern" definition that can be used as a foundation for development of biosecurity screening standards and policy.

bioinformatics↗

bioRxiv2024

FLIGHTED: Inferring Fitness Landscapes from Noisy High-Throughput Experimental Data

Machine learning (ML) for protein design requires large protein fitness datasets generated by high-throughput experiments for training, fine-tuning, and benchmarking models. However, most models do not account for experimental noise inherent in these datasets, harming model performance and changing model rankings in benchmarking studies. Here we develop FLIGHTED, a Bayesian method of accounting for uncertainty by generating probabilistic fitness landscapes from noisy high-throughput experiments. We demonstrate how FLIGHTED can improve model performance on two categories of experiments: single-step selection assays, such as phage display and SELEX, and a novel high-throughput assay called DHARMA that ties activity to base editing. We then compare the performance of standard machine-learning models on fitness landscapes generated with and without FLIGHTED. Accounting for noise significantly improves model performance, especially of CNN architectures, and changes relative rankings on numerous common benchmarks. Based on our new benchmarking with FLIGHTED, data size, not model scale, currently appears to be limiting the performance of protein fitness models, and the choice of top model architecture matters more than the protein language model embedding. Collectively, our results indicate that FLIGHTED can be applied to any high-throughput assay and any machine learning model, making it straightforward for protein designers to account for experimental noise when modeling protein fitness.

bioinformatics↗

bioRxiv2023

PyLabRobot: An Open-Source, Hardware Agnostic Interface for Liquid-Handling Robots and Accessories

Liquid handling robots are often limited by proprietary programming interfaces that are only compatible with a single type of robot and operating system, restricting method sharing and slowing development. Here we present PyLabRobot, an open-source, cross-platform Python interface capable of programming diverse liquid-handling robots, including Hamilton STARs, Tecan EVOs, and Opentron OT-2s. PyLabRobot provides a universal set of commands and representations for deck layout and labware, enabling the control of diverse accessory devices. The interface is extensible and can work with any robot that manipulates liquids within a Cartesian coordinate system. We validated the system through unit tests and several application demonstrations, including a browser-based simulator, a position calibration tool, and a path-teaching tool for complex movements. PyLabRobot provides a flexible, open, and collaborative programming environment for laboratory automation. Figure Abstract O_FIG O_LINKSMALLFIG WIDTH=124 HEIGHT=200 SRC="FIGDIR/small/547733v2_ufig1.gif" ALT="Figure 1"> View larger version (34K): org.highwire.dtl.DTLVardef@9dbd4forg.highwire.dtl.DTLVardef@1f7cc6dorg.highwire.dtl.DTLVardef@18e0d37org.highwire.dtl.DTLVardef@105cd0b_HPS_FORMAT_FIGEXP M_FIG C_FIG PyLabRobot overcomes the limitations of proprietary robotic systems. (a) Scientists with access to liquid-handling robots are currently limited by proprietary interfaces that require specialized knowledge, hinder cross-platform operability, and restrict sharing of methods among different robot types. For complex tasks, many researchers need assistance from a specialist familiar with their particular system, most notably when creating or editing protocols. (b) PyLabRobot (https://github.com/PyLabRobot/pylabrobot) offers a single interface that allows any person with basic Python skills to program diverse types of liquid-handling robots and share protocols freely, fostering a more collaborative and efficient research environment. The Python API makes it easy to interact with a large scientific computing ecosystem and allows users to leverage large language models for programming assistance.

bioengineering↗

bioRxiv2023

A framework for identifying fertility gene targets for mammalian pest control

O_LIFertility-targeted gene drives have been proposed as an ethical genetic approach for managing wild populations of vertebrate pests for public health and conservation benefit. C_LIO_LIThis manuscript introduces a framework to identify and evaluate target gene suitability based on biological gene function, gene expression, and results from mouse knockout models. C_LIO_LIThis framework identified 16 genes essential for male fertility and 12 genes important for female fertility that may be feasible targets for mammalian gene drives and other non-drive genetic pest control technology. Further, a comparative genomics analysis demonstrates the conservation of the identified genes across several globally significant invasive mammals. C_LIO_LIIn addition to providing important considerations for identifying candidate genes, our framework and the genes identified in this study may have utility in developing additional pest control tools such as wildlife contraceptives. C_LI

genetics↗

bioRxiv2022

A multiplexed, confinable CRISPR/Cas9 gene drive propagates in caged Aedes aegypti populations

Aedes aegypti, the yellow fever mosquito, is the main vector of several major pathogens including yellow fever, dengue, Zika and chikungunya viruses. Classical mosquito control strategies, mainly utilizing insecticides, have had success in controlling other mosquito vectors in recent years, but are much less useful against Ae. aegypti, and even these methods are threatened by rising insecticide resistance. This has stimulated interest in new mosquito control mechanisms, notably genetic systems such as gene drives. However, the development of CRISPR/Cas9 gene drive systems has faced challenges such as low inheritance biasing rate, the emergence of resistance alleles, and the possibility of spreading beyond the intended population. Here, we test the regulatory sequences from the Ae. aegypti benign gonial cell neoplasm (bgcn) homolog to express Cas9 in the germline to find an expression timing more conducive to homing. We also created a separate multiplexing (targeting multiple different sites within the target gene) sgRNA-expressing homing cassette inserted into the Ae. aegypti kynurenine 3-monooxygenase (kmo) gene to limit the consequences of resistance alleles. This creates a split gene drive such that one part does not drive, allowing control over geographic spread and temporal persistence. When combined, these two elements provide highly effective germline cutting at the kmo locus and act as a gene drive. Our target genetic element was driven through a cage trial population such that carrier frequency of the element increased from 50% to up to 89% of the population despite significant fitness costs to kmo insertions. Deep sequencing suggests that the multiplexing design could mitigate resistance allele formation in our gene drive system. Significance statementMosquito-borne diseases affect millions of people worldwide, with the yellow fever mosquito (Aedes aegypti) being the principal vector of many viral diseases. Effective measures for controlling this mosquito are sorely needed. Gene drive systems have arisen as a potential tool for mosquito control due to their ability of biasing inheritance of a trait into a target population. Here, we assess a split gene drive, based on CRISPR/Cas9 endonuclease technology driving a target element into the mosquito population. Evaluated over successive generations in a replicated cage trial, the drive successfully biased its inheritance, increasing in frequency from 50% to up to 89%. Our results are encouraging for the potential use of this type of contained gene drive system for mosquito control in endemic areas.

synthetic biology↗

bioRxiv2022

High-throughput molecular recording can determine the identity and biological activity of sequences within single cells

Large datasets of biomolecular activities are crucial for protein engineering, yet their scarcity due to limited experimental throughput hampers progress. We introduce Direct High-throughput Activity Recording and Measurement Assay (DHARMA), an innovative method enabling ultra-high-throughput measurement of biomolecular activities. DHARMA employs molecular recording techniques to link activity directly to editing rates of DNA segments contiguous with the coding sequence of biomolecule of interest. Leveraging a Bayesian inference-based denoising model, we mapped the fitness landscape of TEV protease across 160,000 variants. Using these datasets, we benchmarked popular protein models and showed the impact of data size on model performance. We also developed circuit self-optimization strategies and demonstrated DHARMAs capability to measure a wide range of biomolecular activities. DHARMA represents a leap forward, offering the machine learning community unparalleled datasets for accurate protein fitness prediction and enhancing our understanding of sequence-to-function relationships. O_FIG O_LINKSMALLFIG WIDTH=200 HEIGHT=78 SRC="FIGDIR/small/483646v4_ufig1.gif" ALT="Figure 1"> View larger version (22K): org.highwire.dtl.DTLVardef@47425borg.highwire.dtl.DTLVardef@557a0corg.highwire.dtl.DTLVardef@1bfc091org.highwire.dtl.DTLVardef@1fb7fdf_HPS_FORMAT_FIGEXP M_FIG C_FIG

synthetic biology↗

bioRxiv2020

A high-throughput platform for feedback-controlled directed evolution.

Continuous directed evolution rapidly implements cycles of mutagenesis, selection, and replication to accelerate protein engineering. However, individual experiments are typically cumbersome, reagent-intensive, and require manual readjustment, limiting the number of evolutionary trajectories that can be explored. We report the design and validation of Phage-and-Robotics-Assisted Near-Continuous Evolution (PRANCE), an automation platform for the continuous directed evolution of biomolecules that enables real-time activitydependent reporter and absorbance monitoring of up to 96 parallel evolution experiments. We use this platform to characterize the evolutionary stochasticity of T7 RNA polymerase evolution, conserve precious reagents with miniaturized evolution volumes during evolution of aminoacyl-tRNA synthetases, and perform a massively parallel evolution of diverse candidate quadruplet tRNAs. Finally, we implement a feedback control system that autonomously modifies the selection strength in response to real-time fitness measurements. By addressing many of the limitations of previous methods within a single platform, PRANCE simultaneously enables multiplexed, miniaturized, and feedback-controlled continuous directed evolution.

bioengineering↗

bioRxiv2017

Daisy quorum drives for the genetic restoration of wild populations

An ideal gene drive system to alter wild populations would 1) exclusively affect organisms within the political boundaries of consenting communities, and 2) be capable of restoring any engineered population to its original genetic state. Here we describe daisy quorum drive systems that meet these criteria by combining daisy drive with underdominance. A daisy quorum drive system is predicted to spread through a population until all of its daisy elements have been lost, at which point its fitness becomes frequency-dependent: mostly altered populations become fixed for the desired change, while engineered genes at low frequency are swiftly eliminated by natural selection. The result is an engineered population surrounded by wild-type organisms with limited mixing at the boundary. Releasing large numbers of wild-type organisms or a few bearing a population suppression element can reduce the engineered population below the quorum, triggering elimination of all engineered sequences. In principle, the technology can restore any drive-amenable population carrying engineered genes to wild-type genetics. Daisy quorum systems may enable efficient, community-supported, and genetically reversible ecological engineering.\n\nSummaryLocal communities should be able to control their own environments without forcing those choices on others. Ideally, each community could reversibly alter local wild organisms in ways that cannot spread beyond their own boundaries, and any engineered population could be restored to its original genetic state. We've invented a 'daisy quorum' drive system that appears to meet these criteria.\n\n\"Daisy\" refers to a daisy drive, which typically uses a daisy-chain of linked genes to spread a change through a local population while losing links every generation until it stops spreading. \"Quorum\" reflects the system's ability to \"vote\" on whether a local population should be altered or not: once all daisy elements are lost, it favors replication by the altered version or the original depending on which is more abundant in the local area. Put together, they result in a change that first spreads through a local population, then either becomes locally prevalent is eliminating, inhibiting mixing at the boundary. All organisms in the target population are altered, but changes are unable to spread much beyond that area due to being greatly outnumbered by wild-type organisms and consequently less able to replicate.\n\nWe haven't yet performed any experiments involving daisy quorum systems. Rather, were describing what we intend to do, including the safeguards we will use and our assessment of risks, in the hope that others will evaluate our plans and tell us if there's anything wrong that we missed. We hope that all researchers working on gene drive systems - and other technologies that could impact the shared environment - will similarly pre-register their plans. Sharing plans can reduce needless duplication, accelerate progress, and make the proposed work safer for everyone.

synthetic biology↗