Extended Data Fig. 1: TotalX experimental workflow and analysis pipeline.

a. Schematic overview of the TotalX workflow. Total RNA is enzymatically polyadenylated and reverse transcribed using a custom template-switching oligo (TSO). The TSO is then selectively removed, followed by Cas9-based rRNA depletion (DASH) at the cDNA level. Resulting cDNA fragments are size-separated, with short (< 400 bp) and long (> 400 bp) fractions indexed independently and pooled for sequencing. An optional gel-purified miRNA fraction (~ 18–50 bp) enables targeted enrichment of mature small RNAs. b. Structure of the short-fragment indexed library. The short-fragment indexed libraries can optionally undergo Pippin gel selection system to obtain miRNA-enriched libraries. c. Workflow for combining short RNA, miRNA-enriched and standard cDNA fractions. The resulting composite library captures a broad range of transcript sizes and biotypes. d. Cell Ranger-based data analysis pipeline of TotalX. e. Examples of gene coverage profiles for representative miRNAs located within the lncRNA locus (top), intron of a protein coding gene (middle) and exon of a protein coding gene (bottom). Bam file used for the coverage plots was generated with the modified Cell Ranger pipeline. f. Number of cells per dataset for 10x Genomics, VASA-seq, and TotalX libraries. g. Relationship between UMIs per cell and number of genes detected per cell across datasets and library configurations. h. Per-cell UMI comparison between TotalX and VASA-drop across RNA biotypes. i. Relationship between per-gene mean expression and CV² for 10×3′, VASA-drop, and TotalX. All methods show a Poisson-like regime at low mean (slope ≈ −1) and CV² saturation at higher mean. j. Negative-binomial overdispersion (ϕ) across RNA biotypes for 10×3′, VASA-drop and TotalX. Non-coding RNAs show higher ϕ than protein-coding genes in both methods, but TotalX and VASA-drop display similar overdispersion distributions, indicating comparable technical noise.