Selection criteria for nanobody

Hi,

I have two queries. If someone could help me find the solution, it would be great.

I am working on camelid VHH sequences. I've pre-processed the NGS raw reads (QC filter, merge PE reads, de-duplicate). From the filtered dataset of 1000s of nanobodies (VHH), I want to select a few top 20-30 high-affinity functional nanobodies using a computational approach first, then validate by experimentation. What should be the criteria for such selection? How can I select top candidate sequences? If I consider homology modeling and molecular docking, it seems impractical to do this on 1000s of sequences. Plus, since I am a newbie in this field, so using ML algorithms to predict affinities of unknown sequences is not really my cup of tea.

From a fatsq file of VHH and VH sequences, I want to filter out VHH sequences to use for downstream analysis. There are hallmark residues of VHH that could be used as a parameter to filter VHH, but this parameter alone is not enough. I want to use the additional cysteine motif (another hallmark of VHH only) other than the conserved, which also lies in VH, as a supporting parameter to make the filtration more valid. But I don't know the motif in the first place. Does anyone know the hallmark cysteine motif of VHH?

Thanking in advance.

nanobody

• 2,454 views • link

updated 4 months ago by Fran •

• written 2.4 years ago by farahnazir2 •

For anyone who was wondering what nanobodies are:

Nanobodies are the smallest functional single-domain antibodies known to be able to bind stably to antigens, with the advantages of high stability, high hydrophilicity, and easy expression and modification.

• 0 views • link

written 2.4 years ago by GenoMax

Beyond abundance and clustering, I'd recommend adding an immunogenicity and developability filter before committing to affinity assays — it can save significant wet-lab cost by discarding candidates with high predicted ADA risk or poor biophysical properties early.

We built a free tool for exactly this: ImmunoCheck (https://jfinnova.es/immunocheck). It takes a VHH sequence and returns:

MHC-II epitope density and hotspot positions Humanization grade (comparison against 26 IGHV germlines) Developability profile (aggregation hotspots, chemical liabilities, stability) Overall immunogenicity risk score (0-10) No API key required. If you have thousands of sequences, the API supports batch processing: pip install immunocheck

Could be a useful additional selection criterion alongside your clustering approach.

• 0 views • link

written 4 months ago by Fran •

1 answer

I am working on camelid VHH sequences. I've pre-processed the NGS raw reads (QC filter, merge PE reads, de-duplicate)

You really need to better describe the experiment that generated these reads. Clearly you've got some kind of high-throughput screen that uses DNA sequencing as a readout; but it's totally unclear the nature of the assay and what a "background" would look like, and I can think of many different approaches. Until this is clear, nobody can help you.

• 0 views • link

written 2.4 years ago by LChart

Yes, you're right. Thank you for asking. I should have mentioned the background. A camel was immunized with various antigens (unknown). From the generated immune response, VHH sequences were amplified (by PCR) and sequenced by Illumina PE300. It generated millions of reads. I want to isolate only a few high-affinity functional sequences.

• 0 views • link

written 2.4 years ago by farahnazir2 •

Are you sure there's not a step missing? This approach does not necessarily select for high-affinity VHH sequences. One could assume that there is selection for high-affinity VHH sequences, but I think this is a bad assumption. Typically some kind of assay (ELISA/Phage Display) would be used to further investigate affinities -- e.g. by cloning the sequences into phages for direct binding, or hybridizing the nanobodies to a nucleic acid barcode.

Right now it looks like you're at step 1 which is "what VHH did the camel produce". This can be done with standard assembly software which will produce contigs for each (major) rearrangement. You can then translate into protein sequences and use molecular docking to estimate affinities. However, I think typically the VHH sequences would be directly cloned -- even without sequencing -- into a screening assay, which is why this is a little bit confusing. You're sure this sequencing came straight from the inoculated animal, and not an affinity enrichment assay?

• 0 views • link

written 2.4 years ago by LChart

I think there is a loophole in my experimental design since I have no expertise in this domain. Your questions clear various aspects of the study in my mind, and your valuable suggestions give me a valid direction. Thank you for that first. Secondly, given that I have PE300 reads, the entire VHH sequence fits within the overlap of the paired reads, so only merging could indeed generate a complete sequence without the need for assembly (in my opinion); that's why I didn't generate assembly.

From a dataset of 1000s of translated VHH sequences, if I need to select only a few sequences to test their affinity (using docking), what should be the criteria for choosing those sequences? Selecting highly abundant sequences may overlook the chance of novel transcripts, which may not be that abundant. Can you please suggest something for this?

• 0 views • link

written 2.3 years ago by farahnazir2 •

The only information that is available to you for selection (without using docking itself) is the abundance of each clone, and the sequence divergence between clones. You can likely arrange the VHH sequences into clades via similarity clustering, and evaluate both the clade abundance and within-clade abundance to help select sequences to check.

• 0 views • link

written 2.3 years ago by LChart

Thank you f or providing your valuable suggestions.

• 0 views • link

written 2.3 years ago by farahnazir2 •