Hi, authors

Thanks for your incredible contributions and efforts in this amazing work.

After looking into the code and the Xenium samples downloaded from HF, I found the following issues:

• About spot size in pooling: There are 65 Xenium samples in HEST. They are processed into pseudo-visium ST data by using pool_transcripts_xenium(transcript_df, dict['pixel_size_um_estimated'], key_x='he_x', key_y='he_y', spot_size_um=spot_size_um), producing the ST file st/{sample_id}.h5ad. I checked each ST file and calculated pixel_distance_between_two_adjacent_spots * meta_info['pixel_size_um_estimated'] using the spatial coordinates in the ST file. I found that the value is 100um for most Xenium samples. But for two samples,TENX122 and NCBI784, the value is significantly different from 100um. Concretely, for NCBI784, pixel_distance_between_two_adjacent_spots = 274.784593593837 (derived from st/{sample_id}.h5ad) and meta_info['pixel_size_um_estimated'] = 0.2125 (from HEST_v1_1_0.csv), leading to spot_size_um = 58.39. So, I want to ask, is spot_size_um set to 100um for all Xenium samples when applying pooling to transcripts/{sample_id}_transcripts.parquet? Is there any consideration for setting spot_size_um=100 for Xenium samples?
• Inclusion criteria for Xenium samples in benchmark datasets: For example, for task 1 (oncotree=IDC), 4 samples are included in this task, i.e., TENX95, TENX99, NCBI783, and NCBI785. However, in HEST_v1_1_0.csv, there are 7 Xenium samples: TENX99, TENX98, TENX97, TENX95, NCBI785, NCBI784, NCBI783. Three samples are excluded from the benchmark datasets. It is also the same case for task 4 (oncotree=SKCM). Could you clarify the inclusion criteria for Xenium samples in the benchmark datasets a bit more?

Looking forward to hearing from you. Thanks

Best,
Pei