henhao
Sign in
GenomicsbioRxiv
Preprint

CRISPR-based surveillance for COVID-19 using genomically-comprehensive machine learning design

Hayden C. MetskyiD, Catherine A. FreijeiD, Tinna-Sólveig F. Kosoko-ThoroddseniD, Pardis Christine SabetiiD, Cameron MyhrvoldiD

PDFDOIData
Post to X
Threads on XPost to BlueskyThreads on Bluesky
In-depth full-text digest available · Jump to digest ↓

Abstract

The emergence and outbreak of SARS-CoV-2, the causative agent of COVID-19, has rapidly become a global concern and has highlighted the need for fast, sensitive, and specific tools to surveil circulating viruses. Here we provide assay designs and experimental resources, for use with CRISPR-based nucleic acid detection, that could be valuable for ongoing surveillance. We provide assay designs for detection of 67 viral species and subspecies, including: SARS-CoV-2, phylogenetically-related viruses, and viruses with similar clinical presentation. The designs are outputs of algorithms that we are developing for rapidly designing nucleic acid detection assays that are comprehensive across genomic diversity and predicted to be highly sensitive and specific. Of our design set, we experimentally screened 4 SARS-CoV-2 designs with a CRISPR-Cas13 detection system and then extensively tested the highest-performing SARS-CoV-2 assay. We demonstrate the sensitivity and speed of this assay using synthetic targets with fluorescent and lateral flow detection. Moreover, our provided protocol can be extended for testing the other 66 provided designs. Assay designs are available at https://adapt.sabetilab.org/.

02 /Specification & Classification

Click any tag to browse matching papers
Citations
162
OpenAlex Topics

In-depth summary

✦ Multi-agent verified full-text synthesis

Claude Sonnet 4.6 · max effort · 8 Nov 2026 · from the full text of version 2 (2 Mar 2020) · CC BY-NC-ND 4.0

A machine-learning design pipeline produced CRISPR-Cas13 detection assays for 67 viral taxa; the best of 4 screened SARS-CoV-2 designs reached 10 copies/μl on synthetic RNA.

SARS-CoV-2 diagnostics must be fast and specific, because the virus is closely related to other coronaviruses and patients may carry other respiratory viruses or co-infections. This work supplies pre-designed CRISPR assays across 67 species and subspecies, plus one experimentally characterized SARS-CoV-2 assay.

Key results

  1. ShownThe ADAPT design system generated 67 assays covering SARS-CoV-2, related SARS-like viruses, other Coronaviridae and common respiratory viruses. They are intended for single-target use or grouping into multiplex panels.

    67 assays; comprehensiveness of known sequence diversity >97% for most assays; designs constrained by predicted sensitivity (LwaCas13a only) and predicted specificity. · Abstract; Designs for single assay and multiplex panels; Tables

  2. SuggestedOf 4 SARS-CoV-2 designs screened with a Cas13 SHERLOCK system, the best performer was also the design ranked highest a priori. This agreement is consistent with the ranking being useful, though the screen was small.

    4 designs screened; the top-performing assay matched the highest a priori ranked design (no per-design signal values given in text). · Abstract; SARS-CoV-2 SHERLOCK assay testing

  3. ShownThe selected SARS-CoV-2 assay detected synthetic RNA at a limit of detection of 10 copies/μl with both fluorescent and lateral flow readouts. The authors state it performs well relative to the published DETECTR and SHERLOCK assays.

    LoD 10 copies/μl (both readouts) vs DETECTR 70-300 cp/μl and SHERLOCK 10-100 cp/μl; fluorescence error bars are 1 SD over n=3 technical replicates. · Introduction; SARS-CoV-2 SHERLOCK assay testing; Figure legends

  4. ClaimedPredicted sensitivity and specificity of most designs, and specificity of the SARS-CoV-2 assay against related viruses, have not been tested experimentally. The authors present these as supported by prediction and in silico checks.

    Most designs untested; specificity testing against related viruses and clinical or patient samples listed as planned work; in silico verification covers comprehensiveness and partly specificity. · Introduction; Designs for single assay and multiplex panels; Discussion

Methods, data and reproducibility

ADAPT algorithms and machine learning models design assays under constraints of coverage of known genomic diversity, predicted Cas13 activity and predicted specificity. Four SARS-CoV-2 designs were screened, then the best was tested on synthetic RNA with fluorescent and lateral flow readouts.

  • ADAPT-designed CRISPR-Cas13 assays for species and subspecies of coronaviruses and other respiratory viruses · virus (SARS-CoV-2, other Coronaviridae, respiratory viruses) · n = 67 designs
  • SARS-CoV-2 SHERLOCK assay candidates screened experimentally · SARS-CoV-2 · n = 4 designs
  • Synthetic SARS-CoV-2 RNA target dilution series for limit of detection testing (fluorescent and lateral flow) · SARS-CoV-2 (synthetic RNA) · n = n=3 technical replicates (fluorescence)
Methods & tools
ADAPT (assay design system), LwaCas13a SHERLOCK detection, Recombinase polymerase amplification (TwistAmp Basic), RevertAid Reverse Transcriptase, T7 RNA polymerase, RNase Alert v2 fluorescent reporter, HybriDetect lateral flow strips, GISAID genome data
Data & accessions
https://adapt.sabetilab.org/
Reagents & environment
Synthetic SARS-CoV-2 RNA targets; LwaCas13a with crRNA; RPA with reverse transcriptase; fluorescent (RNase Alert v2) and lateral flow (HybriDetect, Milenia) reporters. Primer, target and crRNA sequences are in Supplementary Table 1.
Experimental conditions
Amplification at 41 °C for 20 minutes; Cas13 detection at 37 °C for 30 minutes to 3 hours, fluorescence read every 5 minutes (qPCR machine or plate reader); lateral flow strips read after 2-5 minutes.
Reproducibility gaps
The ADAPT algorithms are described as a manuscript in preparation; no code repository is given. Design outputs are hosted on a website; most designs are untested.
Limitations (5)
  • Most of the 67 designs have not been experimentally tested; experimental work focused on a single SARS-CoV-2 assay. · stated by the authors
  • The SARS-CoV-2 assay has been tested only on synthetic RNA; sensitivity on clinical isolates and patient samples (sputum, throat, nasal swabs) is still to be evaluated. · stated by the authors
  • Specificity against highly related viruses at species and subspecies levels remains to be tested experimentally with mixtures of synthetic targets and patient samples. · stated by the authors
  • Predicted sensitivity of the designs is based on LwaCas13a activity only. · stated by the authors
  • Comparison with DETECTR and SHERLOCK sensitivities is against previously disclosed values, not a head-to-head experiment; error bars derive from n=3 technical replicates only. · noted in this summary

Who should read:Read this if you need ready-made CRISPR-Cas13 assay designs for SARS-CoV-2 and related respiratory viruses, along with a first synthetic-RNA validation of one assay.

Reproducibility ledger

Nobody has reported trying to reproduce this paper yet. Did you replicate an experiment or run the code? Your report saves the next lab a week.

Sign in to file a reproducibility report.

Related papers

Similarity Map
SelectedSame fieldCross-field
THISGuo '20Curti '20Ding '20Ansari '23Mann '22Ooi '20
75% semantic matchMolecular Biology
Lu Guo, Xuehan Sun, Xinge Wang+13 more·Cell Discovery·May 18, 2020· 246 citations