Challenges of differentiating close homologs within a gene family
Genes in a family are hard to differentiate due to high sequence homology, making it challenging to design probes that will bind exclusively to one member of the family across all targeted, probe-based technologies (e.g. Vizgen, CosMx, Visium, etc.). There is a design tradeoff between being able to detect the expression of a gene, vs. ensuring that the probe does not bind to a closely-related variant of that gene. As the number of targeting probes for a gene goes up, the ability to discriminate between close variants goes down.
Gene family analysis with Xenium
Probes in the Xenium panels are designed to minimize off-target binding across the transcriptome and low FDR counts in our assays validate the effectiveness of our panel design algorithms. During the panel design process, tradeoffs between sensitivity and crossbinding to homologs of a gene family are made to optimize between detection of a gene family vs. discriminating between gene family members.
Our off-the-shelf (OTS) predesigned panels were not specifically designed to distinguish closely related gene family members for all genes. If the experimental goal is spatial cell-typing and clustering, then the existing panels are fully capable of answering those questions.
On the other hand, if the experimental goal is to study the differential expression of closely related genes from a gene family with,
- Predesigned on-market panels
This is not always possible for current, on-market Xenium v1 predesigned panels because while the probes are designed to minimize binding between homologs, they are not optimized for differentiating different homologs in a gene family. For 5k Prime predesigned panels, each probeset has a distinct codeword assignment and customers can reach out to 10x Applied Bioinformatics to obtain an updated JSON file where probesets that may bind to another homolog (that is not of interest to the customer) are removed from the file for analysis.
- Custom panels
For both Prime and v1 panels, customers will need to initiate an advanced panel design process and communicate with the 10x Applied Bioinformatics team the sets of homologs that need to be differentiated with high specificity in the panel. The 10x Applied Bioinformatics team will assess and design probes that have higher probability of differentiating homologs of interest across a gene family.
Gene family analysis strategies for existing data
For v1 and Prime chemistry, it’s worth comparing Xenium data against single-cell data from similar samples. As an example, suppose VarA and VarB are two genes from the same family. If single-cell data suggests that a particular cell type only expresses VarA, then one possibility is to ignore low counts of VarB from the Xenium data for the same cell type, since those probes might be binding to VarA transcripts.
For Prime chemistry, it is also possible to email support@10xgenomics.com and work with 10x Applied Bioinformatics to obtain an updated JSON file where the probesets that may bind to another homolog (that is not of interest to the customer) are removed from the file for analysis. Customers can then run xeniumranger relabel using the new JSON file to get new outputs.
Additional information
List of potential gene family homologs in Xenium OTS panels (link to CSV file):