Inqiry on Splicing Junction Visualization and Data Extraction

Help & Support
1

Hi everyone! I’m relatively new to coding and have been using AlphaGenome to visualize predicted splicing junctions with Sashimi plots. My main goal is to explore how different genes express tissue-specific isoforms due to alternative splicing.

During my analysis, I ran into a few hurdles and would love some guidance from the community:

  • Missing Spleen Plot: When trying to generate a Sashimi plot for CHRM3 splicing junctions across the pancreas, ovary, and spleen (using the filter_by_tissue function), the spleen plot failed to generate. Is this likely a syntax/coding error on my end, or is there a known limitation in AlphaGenome’s predictions for certain organs?

  • Extracting Sashimi Plot Scores: I know that raw splicing junction data can be extracted, but is it possible to simultaneously export the specific scores displayed on the Sashimi plots?

  • Approach Validation: Lastly, I just want to confirm if using AlphaGenome to compare cross-tissue isoform variations is a valid and robust approach for this type of research.

Much appreciation for any help or insights you can share!

2

Hi There!

Thanks for reaching out.

I’ve checked our available biosample/modality pairings and you should be able to predict slice junctions for UBERON:0002106 (spleen). The filter_by_tissue function filters based on the ‘gtex_tissue’ column which can be found in output.reference.splice_junctions. I can only see one option here (“Spleen”); make sure spelling is correct. If no junctions are predicted then the arc chart will not display. Can you share a code snippet so I can check?

Yes it should be possible to calculate the arc values based on the values stored in JunctionData.values. Bear in mind that these values will undergo normalisation in order to be plotted, and only those values exceeding a plotting threshold will be visible above arcs. More details on normalisation and filtering can be found on our github page here.

sashimi_component = plot_components.Sashimi(
            output.reference.splice_junctions
        )
# Access the underlying track from the Sashimi component
junction_track = sashimi_component._junction_track

# Now extract the numerical array and metadata
raw_values = junction_track.values
metadata = junction_track.metadata

print("Shape of junction values:", raw_values.shape)
print("Metadata details:\n", metadata)
print(raw_values)

Kind regards,
Nicolene