SENTINEL - Smart Environmental Nucleic-acid Tracking using Inference from Neural-networks for Early-warning Localization
Important
This is an educational repository for installing ADAPT dependencies for CRISPR-based environmental biosurveillance (CRISPR-eBx) deployments, such as SENTINEL education (SENTINEL-ed). Many things in this repository have been adapted from the Original ADAPT repository to enhance educational accessibility. The original repository is recommended if you want to develop ADAPT further. There is a copy of the ADAPT v1.6.0 repository in this repo if ever required.
Note
Our SENTINEL article has been accepted at the Environmental DNA journal, where all this pipeline and similar procedures were done.
You can explore our article at Durán-Vinet et al. CRISPR-based environmental biosurveillance assisted via artificial intelligence design of guide-RNAs (2025)
Southern Environmental DNA Society (SeDNAs)
Consortium for the Application of CRISPR in Ecology (CACE)
Get in touch! You can contact GO, seDNAs or directly to benjamin.duran-vinet@postgrad.otago.ac.nz
- SENTINEL introduction
- Quick glossary
- Before starting
- Installing ADAPT
In particular, SENTINEL is an integrated tool for CRISPR-based environmental biosurveillance (CRISPR-eBx). SENTINEL currently leverages ADAPT (Activity-informed Design with All-inclusive Patrolling of Targets) [article here]. ADAPT is an end-to-end trained artificial intelligence with ~19,000 guide-target pairs, ultimately providing a robust and comprehensive platform for target discovery in the environmental nucleic acid field that can accelerate environmental biosurveillance of emergent biosecurity threats or aid endangered species detection. ADAPT designs are trained to be highly sensitive and specific, with well-established command options that enable high customization for a flexible assay design for the user. ADAPT designs are also scalable and can be locally deployed.
This ADAPT platform will provide suitable ranked guide-target pairs for the given target (set of primers and a spacer sequence, see below).
Illustration from Durán-Vinet et al. (in press).
The ultimate objective of SENTINEL is to streamline and accelerate CRISPR-eBx deployments, whether from air, soil or water samples. So far, only water has been tested, but air and soil environmental samples are a promising deployment.
Most CRISPR-eBx deployments use an isothermal amplification method: RPA (recombinase polymerase amplification) and LAMP (Loop-mediated isothermal amplification). A quick illustration is shown below for RPA-CRISPR-Cas13a.
Illustration from Durán-Vinet et al. (in press).
Important
Reading this glossary will enhance the overall experience when going through the workshop and the manual.
CRISPR-eBx, CRISPR-based environmental biosurveillance, an alternative use of CRISPR to detect target species presence from environmental samples, targeting environmental nucleic acids. This could be for various applications, including environmental viral vectors, elusive endangered species, invasive species, diseases, genetic modifications and more.
CRISPR-Dx, CRISPR-based diagnostics, CRISPR-Dx is the original use of CRISPR-based detection tools. However, these clinical deployments are not considered the same as an environmental deployment, i.e., CRISPR-eBx.
DR, Direct repeat, section of the gRNA specific for each Cas nuclease.
eNAs, Environmental nucleic acids, often refer to all the DNA or RNA shredded or left behind in the environment by organisms or agents. eNAs include environmental DNA (eDNA) and environmental RNA (eRNA).
Guide-target pairs, these are defined as the construct of primers and specific gRNAs that will detect a certain genomic region.
gRNA, small RNA molecules that program and guide the Cas nuclease into a specific target by nucleic acid complementarity. These are also sometimes called CRISPR RNAs (crRNAs).
LAMP, Loop-mediated isothermal amplification. One of the most used isothermal amplification methods, it uses around 4 - 6 primers and runs between 60-70°C.
RPA, Recombinase polymerase amplification. One of the most used isothermal amplification methods uses 2 primers and can run between 25-42°C.
Some noteworthy deployments of CRISPR-based environmental biosurveillance are:
- Biomonitoring of endangered fish Williams et al., 2019 - The application of CRISPR-Cas for single species identification from environmental DNA in Mol. Ecol. Res.
- Biomonitoring of endangered amphibians Leugger et al., 2024 - Scanning amplicons with CRISPR-Dx detects endangered amphibians in environmental DNA in Mol. Ecol. Res.
- Biomonitoring of marine invasive species
- Quick, in-field biosurveillance Yang et al., 2024 - Rapid, easy, sensitive, low-cost and on-site detection of environmental DNA and RNA using CRISPR-Cas13 in Methods Ecol. Vol.
Some noteworthy reviews and perspectives on the CRISPR-based detection field are:
- CRISPR-based diagnostics for harmful algal blooms Durán-Vinet et al., 2021 - Potential applications of CRISPR/Cas for next-generation biomonitoring of harmful algae blooms: A review in Harmful Algae
- CRISPR-based diagnostics Kaminsky et al., 2021 - CRISPR-based diagnostics in Nat. Biomed. Eng.
- CRISPR-based diagnostics integrated with deep learning Durán-Vinet et al., 2023 - CRISPR-Cas-Based Biomonitoring for Marine Environments: Toward CRISPR RNA Design Optimization Via Deep Learning in CRISPR J.
- CRISPR-base depletion Kardailsky et al., 2025 - Monitoring the Land and Sea: Enhancing Efficiency Through CRISPR-Cas Driven Depletion and Enrichment of Environmental DNA in CRISPR J.
This repository is educational and is not intended to cover all of bioinformatics. Instead, it is intended to teach you to use a specific tool within many other applications that Conda could run. If you ran into any potential issues when exploring ADAPT deeper, feel free to contact the original ADAPT authors or me at benjamin.duran-vinet@postgrad.otago.ac.nz
Important
The dependencies listed below are automatically installed via Bioconda, and they differ from the original ADAPT. These changes has been tested and do not change the results quality, but allow better compatibility across systems. If you desire to install the original package, it has to be installed via pip or by downloading the repository, but this is ONLY for advanced/dev users.
Moreover, the original version of ADAPT can only be installed in Linux or Windows due to major incompatibilities with the arm64 architecture with Protobuf.
ADAPT will install:
- Python == 3.8.18
- NumPy == 1.24.3
- SciPy == 1.10.1
- TensorFlow == 2.11.0
Using the thermodynamic modules of ADAPT requires:
- Primer3-py == 0.6.1
To run ADAPT, you will need conda installed on your Windows Linux Subsystem or macOS. Please make sure you install it before installing ADAPT; a short walkthrough is included below.
Warning
This is only for Windows users who haven't used Windows Linux Subsystem before.
If you are a Windows 10 user, please follow this Win10 tutorial
If you are a Windows 11 user, please follow this Win11 tutorial
After you have installed Ubuntu subsystem, open the terminal and go to Step 1.1
Important
If you have used the terminal before, you might have conda installed already.
Please check if you already have conda installed
conda --versionThis should provide a number (format xx.x.x). If you got a number as shown below, proceed to step 3 (Creating an environment; Linux and Mac users)
If you got an error, conda is not installed on your computer, then proceed to step 1.2 (Linux subsystem users) or 1.3 (Mac users).
wget https://repo.anaconda.com/miniconda/Miniconda3-latest-Linux-x86_64.sh
bash Miniconda3-latest-Linux-x86_64.shAfter this, restart your terminal. You should have a (base) before you prompt.
Caution
Please ensure you install the correct conda version for Mac, as there are different MacOS architectures (amd64 and arm64); installing the incorrect one will bring downstream compiling errors.
For arm64 architecture for Apple Silicon (i.e., from M1 and onwards):
curl -O https://repo.anaconda.com/miniconda/Miniconda3-latest-MacOSX-amd64.shFor amd64 architecture:
curl -O https://repo.anaconda.com/miniconda/Miniconda3-latest-MacOSX-arm64.shThese commands will download the latest conda version for the correct architecture, Please double-check your architecture before proceeding.
Then run:
bash Miniconda3-latest-MacOSX-*.sh
This will lead you to accept all T&C, press the ENTER key till a yes|no prompt appears.
Type 'yes' and press ENTER. This will install miniconda.
After this, restart your terminal. You should have a (base) before your prompt as shown below:
pwd
'pwd' command allows you to see your current path (where you are in your terminal.
ls
'ls' command allows you to see directories in your current path.
mkdir SENTINELv1
'mkdir' command will create a directory in your current path that you saw with 'pwd'. You can see this new directory using 'ls'.
cd SENTINELv1/
'cd' allows you to go into a specified directory or path. You can use TAB to autocomplete. 'ls' will show that this directory is empty.
cd ..
'cd..' will take you one directory back. You can check with 'pwd' your new position. You will also see that your prompt changed too.
Note
The steps below are for Windows Linux Subsystem, and Mac users, as the command lines are similar.
Tip
Environments are useful tools in bioinformatics as they allow us to have a unique 'bench' for specific work; this way, we avoid creating incompatibilities in our main system files.
conda create -n SENTINELv1 python=3.8
This will create a specific environment that has Python version 3.8.x. It will also install other dependencies and libraries. Type y and press ENTER. This will take some minutes to run.
mkdir SENTINELv1
cd SENTINELv1
mkdir input output
ls
If you have already created a 'SENTINELv1' folder, use the last three lines instead.
Warning
Ensure you activate the environment you created in step 3.1 before continuing.
conda activate SENTINELv1
Your prompt environment should change from (base) to (SENTINELv1). If it activated correctly, you should see Python 3.8. x when using 'conda list'.
conda list
Proceed to install
conda install -c bioconda "adapt[thermo]"Type 'y' and then ENTER. The installation can take some minutes as some dependencies are heavy (mainly TensorFlow). TensorFlow is a software library for machine learning deployments, which is the backbone of ADAPT.
Then:
pip install primer3-py==0.6.1This should have ADAPT 1.6.0 ready to go. Run the command below.
design.py --help
If something similar displays, then ADAPT has been successfully installed. 'design.py' calls for the workflow activation, while '--help' provides specific information to run the workflow. Check out the Commands Explained guide for more details.
Tip
Before starting~~
Always keep a log of your --seed xxx number, this will ensure reproducibility in case you might need it.
You can use Sublime Text for quick command changes and Benchling to keep a trackable record of what you have ran through the terminal.
ADAPT can only read .FASTA files. ADAPT will always assume that input files are aligned unless otherwise specified. It is recommended that files are aligned and examined carefully before running them through ADAPT.
Note
You can optionally use your own FASTA file. In fact, it is encouraged so you can get a useful product from this workshop. You can follow the same instructions provided below, just make sure to use the correct file name.
In this same repository, go to the Example folder and download Test.fasta
Move Test.fasta into your 'Input' directory that was previously created. You can do this manually in Windows or Mac.
To do this, go to your input folder
cd SENTINELv1/input/Tip
If you are a Windows user use:
explorer.exe .Tip
If you are a Mac user use:
open .This will open your current location folder, where you can drag and drop the test.FASTA file and then run 'ls' to see it also in the terminal.
Before running the pipeline, make sure you are in the 'SENTINELv1' folder, as the command shown below has the path directories to work from 'SENTINELv1'. You can call ADAPT from your home folder, but you would need to add the full path so ADAPT can find the input files and deposit the output file properly.
Use 'pwd' to know your current PATH position. You should be in the 'input' folder. If so, use:
cd ..
Then, please copy the following commands and paste them into your terminal.
Warning
Make sure you change your output file name before running a new ADAPT query, or it will get OVERWRITTEN, and previous results will be PERMANENTLY LOST.
Long sequences may require special commands to ensure a smooth run or eventually a server. See Commands Explained for more information.
Ensure you are in the SENTINELv1 folder before running the command below, or an error will occur.
Ensure you have the conda environment activated!
design.py complete-targets fasta ./input/Test.fasta -o ./output/example-output1 --obj maximize-activity --id-m 4 --id-frac 0.01 -gl 28 -gm 0 -pl 30 -pm 5 -pp 0.98 --primer-gc-content-bounds 0.35 0.70 --maximization-algorithm random-greedy --predict-cas13a-activity-model --best-n-targets 10 --seed 001 --verbose'./' fills your path till your actual location in the terminal. The above command will only work if you are in the /SENTINELv1/ folder.
This is how it will look when running
You can see your output file as follows
cd output
ls
Tip
For MacOS users:
open .
Tip
For Windows Linux Subsystem users:
explore.exe .Open the results from Finder (Mac) or Explorer (Windows), but not through the terminal. Results will look similar to those options shown below
OR
To make the results more accessible, if they are open as a text file, copy all the elements and paste them into Excel. This Excel file will be used in Step 5. Alternatively, you can also open it directly as an Excel file.
An explanation of the most important values are given at Understanding Output
Note
To accelerate and enhance the visualisation of results, we recommend using Geneious Prime for the following steps.
Tip
Remember that ADAPT generates specific spacers for LwaCas13a. However, the pipeline can also make PCR and RPA primers for standalone applications.
Warning
ADAPT designs are not ready to use straightaway. We will use Geneious Prime to make them functional.
Spacer and reverse primers (left-primers) have to be reverse-complemented.
Forward primers have to be appended with the T7 promoter.
Spacers must be converted to RNA and appended with a DR sequence to be functional.
Tip
Save the sequences given below in Geneious Prime as separate sequences in a dedicated folder.
See below LwaCas13a DR sequence,
LwaCas13a DR sequence: GAUUUAGACUACCCCAAAAACGAAGGGGACUAAAAC
This sequence has to be appended to the 5' end of the spacer sequence obtained.
See below the T7 promoter,
T7 promoter sequence: GAAATTAATACGACTCACTATAGGG
This sequence has to be appended to the 5' end of the forward primer sequence.
See below the poly-U5 sequence,
polyU5 sequence: /6-FAM/rUrUrUrUrU/BHQ-1/
This is the specific reporter for LwaCas13a. This nuclease has a high semi-specific di-nucleotide motif preference for U-U pairs.
Note
This section is done with Geneious Prime. Make sure to activate your educational license provided.
For the following example, we will use ONLY ONE GUIDE-TARGET PAIR: the best guide-target pair from the previous example, usually the first one of the list in Excel. Copy and paste each sequence independently from that guide-target pair from the Excel file into a dedicated folder in Geneious Prime, and also drag and drop the test.fasta file in the same Geneious Prime folder.
When naming guide-target pairs, it is recommended to use the crRNA position for all IDs.
Note
Remember to create a dedicated folder in Geneious Prime!
After copying and pasting a single guide-target pair in your dedicated Geneious Prime folder, it should look like below:
Warning
As this is a deep learning model, you might have got a different result; proceed with the best predicted target.
Caution
It is recommended to run BLAST on raw results.
Because SENTINEL uses the crRNAs as the main source of specificity while RPA is enrichment only, it is only required to BLAST the spacer sequence. It is also optional to BLAST the primers, but they will have off-targets. Moreover, RPA is highly tolerant of mismatches.
When BLASTing in Geneious, it will create a new folder with full-hit targets, and all hits shown in the list are guaranteed full activity.
As expected, the crRNA has lots of full-hit off-target sequences. This is completely expected as no '--specific-against-fastas' command was used. We will continue with this one as a demonstration. However, if you want ADAPT to provide highly specific guide-target pairs, it is recommended to use '--specific-against-fastas'. You'll need to specify a PATH to the off-target .FASTA file.
Warning
Geneious Prime does not allow bulk 'reverse complement' of primers, MAKE 'reverse complement' before converting the sequences into primers. This is especially useful when working with bulk guide-target pairs.
Reverse complement the Rv and crRNA using the 'reverse complement' option from the Sequence bar.
The sequences will change name automatically with a (reversed) after saving.
Then, select all sequences and convert all guide-target pair sequences into primers using the Primer button.
Now, unselect all and only select your crRNA construct and convert to RNA. Then save.
Now, copy the LwaCas13a DR sequence, go to the crRNA544 sequence, click on 'Allow editing', click on the 5' end and paste the sequence. Repeat the same with the T7 promoter for the Fw primer.
crRNA:
Fw:
As the final stage for all guide-target pairs, a screening has to be done. Select ONLY test.fasta, go to the primer symbol and select 'Test Saved Primers'
Click on 'Choose...' and selected the guide-target pair sequences.
And use the following configuration for your screening, then select OK.
Now, the obtained guide-target pair is successfully screened.
Congratz! You've made it!!!
Note
Use Excel for this optional section.
For massive screenings of results, copying and pasting single guide-target pairs is not feasible. Accordingly, using the same result files in Excel, insert a new full column on the left-side of the columns: 'left-primer-target-sequences'; 'right-primer-target-sequences' and 'guide-target-sequence-positions'. Use the names 'Fw', 'Rv' and 'crRNA', respectively, for these new columns.
Then, type on the cell below Fw:
=("Fw"&W2) and then scroll down the formula. W2 is the crRNA position.
Repeat with Rv and crRNA.
Then, save the Excel file as .CSV file.
Then, drag and drop the .CSV file into a dedicated folder in Geneious Prime. Select .CSV.
Next, select the 'Fw' column as name and the 'left-primer-target-sequences' as sequences, as shown below. Then click on OK.
All Fw primers were inserted in bulk into Geneious Prime, and repeat the same steps but for your Rv and crRNAs sequences. Then you can repeat the Section 5 of this manual.
