Library size can undermine accurate molecular and phenotypic subtyping in spatial transcriptomics data.
Fisher, N. C.; Malla, S. B.; Jamieson, N.; Dunne, P. D.
Show abstract
In an era where transcriptomics-based subtyping, phenotyping and mechanistic understanding is increasingly being driven by state-of-the-art spatially resolved transcriptomic (ST) technologies, it is imperative that researchers, journals, and funders do all they can to ensure that as a community we are interpreting these exciting data as accurately as possible with awareness of their limitations. In this short report, we highlight one potential bias in ST data that could undermine accurate interpretation of transcriptional signatures, providing the field with an opportunity to identify and avoid this issue prior to release of new mechanistic findings. This issue is particularly relevant for platforms that produce some of the most granular and high-resolution spatial information at single cell (and sub-cellular) resolution, with the compromise of a reduced transcriptome panel of genes (Figure 1A). O_FIG O_LINKSMALLFIG WIDTH=141 HEIGHT=200 SRC="FIGDIR/small/602370v1_fig1.gif" ALT="Figure 1"> View larger version (42K): org.highwire.dtl.DTLVardef@1219fb3org.highwire.dtl.DTLVardef@7bd4a4org.highwire.dtl.DTLVardef@1c55c42org.highwire.dtl.DTLVardef@2bf924_HPS_FORMAT_FIGEXP M_FIG O_FLOATNOFigure 1.C_FLOATNO Visualisation of the iCMS3 up signature A: Schematic overview of enrichment using a full gene panel versus a reduced panel. B: Overlap between the genes represented on the CosMx and Xenium gene panels. C: Correlation between the single sample scores of the full iCMS3 up signature (74 genes) and the 15 genes from the signature represented on the CosMx platform. Two samples are highlighted which have a similar iCMS3 up enrichment for the full signature (CRC-JSC-S06: 4535.746; SMC16: 4265.263) but extreme enrichments for the signature composed only of the genes present on the CosMx array (CRC-JSC-S06: 301.0396; SMC16: 11430.593). Median (4047.994) shown by red line. D: Visualisation of the rank of each sample across for the full (CRC-JSC-S06: position 25872/44458; SMC16: position 23907/44458) and CosMx (CRC-JSC-S06: position 513/44458; SMC16: position 44241/44458) signature enrichment. E: Heatmap showing the relative enrichment of each gene with the iCMS3 up signature for each sample, with the genes present on the CosMx array in red. F: Subset of samples (n=628) +/-1% of the median (4007.514 - 4088.474), with the top 100 and bottom 100 samples for CosMx enrichment in red. G: Heatmap of the top 100 and bottom 100 samples (shown in red in F) for CosMx enrichment with a full iCMS3 up enrichment around the median. The samples are arranged by the rank of each sample for CosMx signature enrichment, with the sum of the genes within the iCMS3 up signature present on the array (CosMx [n=16]) and those not present on the array (Non CosMx [n=58]) overlaid as barplots. C_FIG
Matching journals
The top 10 journals account for 50% of the predicted probability mass.
Similar papers in this journal
Similar papers in this journal
- Resolving cellular systems by ultra-sensitive and economical single-cell transcriptome filtering 94%
- Proximity and metabolic activity of the Tumour Microenvironment as predictors of survival in High Grade Serous Ovarian Cancer (HGSOC) 93%
- Network-based approach elucidates critical genes in BRCA subtypes and chemotherapy response in Triple Negative Breast Cancer 92%
"Similar papers" are the closest papers from that journal in the model's embedding space. They show what the match is built on, but the ranking comes mostly from a classifier over the whole training set, not from these examples alone.