Use Nanobody Polyreactivity Online
Commercially Available Nanobody Polyreactivity No-Code Web Server
Nanobody Polyreactivity: Machine Learning Prediction and Reduction of Nanobody Off-Target Binding
The Nanobody Polyreactivity tool is based on an in silico method that predicts how polyreactive a nanobody (camelid VHH) sequence is and suggests substitutions that reduce it. The method was developed by Edward P. Harvey, Andrew C. Kruse and colleagues, including Debora S. Marks, with Kruse at the Blavatnik Institute, Harvard Medical School, and published in Nature Communications in 2022. The authors also provide a companion web server.
Polyreactive antibodies bind off-target molecules. According to the authors, this undermines screening, reproducibility and clinical development of antibodies used as research tools and therapeutics. Polyreactivity is usually measured experimentally and is hard to anticipate from sequence alone.
To build a predictor, the team constructed a naive synthetic nanobody library, sorted it by binding to a polyreactivity reagent made from insect-cell membranes, sequenced the low and high pools, and trained models to classify and score polyreactivity from the CDR sequences. They then used the models to propose substitutions and tested them experimentally.
Attribute | Detail |
|---|---|
Task | Predict nanobody polyreactivity from sequence and rank rescue mutations |
Training data | Yeast-display nanobody library of more than 2 x 10^9 clones, sorted into low- and high-polyreactivity pools |
Models | One-hot and 3-mer logistic regression, CNN and RNN |
Classification | AUC above 0.8 reported for the main models |
Experimental check | Over 90% of predicted substitutions reduced polyreactivity across three nanobody scaffolds |
How the Polyreactivity Models Were Built and Validated
Reagent and library: a polyreactivity reagent was prepared from Sf9 insect-cell membranes and used to sort a yeast surface display library by magnetic and fluorescence-activated cell sorting.
Sequencing: low- and high-polyreactivity pools were sequenced, with CDRs aligned using IMGT numbering.
Model training: position-specific one-hot logistic regression, 3-mer logistic regression, a convolutional network and a recurrent network were trained, with train and test sets split by clustering so test sequences were distinct from training sequences.
Mutation scanning: single and double substitutions were scored in silico to find changes predicted to lower polyreactivity.
Nanobody Polyreactivity Prediction Accuracy and Experimental Results
Model | Classification AUC (initial data) |
|---|---|
One-hot logistic regression | 0.85 |
RNN | 0.84 |
3-mer logistic regression | 0.83 |
CNN | 0.78 |
Isoelectric point alone | 0.79 |
Quantitative scores and rescue mutations
Quantitative ranking: on an index set of nanobodies, predicted scores correlated with measured polyreactivity at Spearman rho of 0.77 (one-hot) and 0.79 (3-mer).
Rescue substitutions: across three scaffolds, 45 of 49 tested substitutions reduced polyreactivity.
Functional nanobody: in an AT1R antagonist nanobody, all 12 tested single substitutions reduced polyreactivity binding, and selected variants kept at least 80% AT1R binding, with preserved radioligand competition and signaling suppression.
Expanded data: deeper sequencing gave about 1.2 million low- and 1.1 million high-polyreactivity clones and similar AUCs of 0.83 to 0.84 for most models.
What is Tamarind Bio?
Tamarind Bio is a no-code bioinformatics platform built to give life scientists and researchers access to powerful computational tools. Many cutting-edge machine learning models are hard to deploy and use. Tamarind provides an intuitive, web-based environment that removes the complexity of high-performance computing, software dependencies and command-line interfaces.
The platform is designed for biologists, chemists and other researchers who may not have a background in programming or cloud infrastructure but want to run models on their own data. Key features include:
A user-friendly graphical interface for setting up and launching experiments
A robust API for integration into existing research pipelines
An automated system for managing and scaling computational resources
Tamarind treats information and data security as a top priority, as detailed in its Trust Center and Terms of Service.
Accelerating Discovery with Nanobody Polyreactivity on Tamarind Bio
Early developability screening: flag polyreactive nanobody candidates before expression and assays.
Rescue of polyreactive leads: rank single substitutions predicted to lower off-target binding while keeping function.
Library triage: score sequences from display campaigns to prioritize cleaner clones.
Directed evolution analysis: check whether mutations that emerge during affinity maturation are predicted to increase or decrease polyreactivity.
How to Use Nanobody Polyreactivity on Tamarind Bio
Open the tool: log in to tamarind.bio and select Nanobody Polyreactivity.
Provide sequences: enter one or more nanobody sequences; the authors' web server accepts FASTA input.
Check numbering if exposed: the method uses IMGT numbering of the CDRs.
Run the job: submit and wait for the predictions.
Download outputs: retrieve the polyreactivity scores and biochemical properties the tool reports.
Interpret: compare scores across candidates; the paper notes isoelectric point and arginine content correlate with polyreactivity.
Test substitutions: validate any suggested changes experimentally, since stability can change.
Things to Keep in Mind
Polyreactivity is assay- and reagent-dependent; the lysozyme ELISA did not correlate with the other assays.
Models were trained on nanobody CDR sequences from one synthetic library; extension to conventional antibodies is not demonstrated.
Functional testing covered a single nanobody, and the quantitative index set was small.
Reducing polyreactivity does not guarantee stability; one substitution slightly lowered thermal stability.
Predictions are computational proxies and do not replace in vivo pharmacokinetic testing.
Source: Harvey et al., An in silico method to assess antibody fragment polyreactivity, Nature Communications 13, 7554 (2022), DOI 10.1038/s41467-022-35276-4. The paper also describes a companion web server.