# 03. Giving Cells and RNA Molecules an Address > How microscopic droplets separate cells and label each RNA copy with its cell of origin and original molecule so copies created during the experiment can be distinguished. Before cDNA, the DNA copy of RNA, from thousands of cells is pooled, each molecule receives two addresses. A **cell barcode** records which cell it came from, while a **unique molecular identifier** (UMI) records which original RNA molecule it came from within that cell. A short sequence fragment produced by the sequencer is called a **read**. > **Question for this lesson**: How can the cell of origin and original molecule count be recovered after all cDNA is pooled? ## A droplet becomes a small workspace Droplet methods such as 10x Chromium partition a cell suspension and barcoded gel beads into tiny droplets. The target is one cell and one bead per occupied droplet. The bead also carries a common sequence for polymerase chain reaction (PCR), which amplifies the small amount of cDNA for analysis. An oligonucleotide is a short DNA strand. Many oligonucleotides are attached to each bead, and each contains four functional parts. | part | role | | --- | --- | | PCR handle | common sequence used to copy cDNA and sequence it later | | cell barcode | shared by cDNA captured in the same droplet | | UMI | a short, varying identifier for an original RNA molecule | | poly(dT) | binds the run of A bases at the end of mRNA and starts copying RNA into cDNA | The cell opens inside the droplet and mRNA binds the bead. Reverse transcription, the process that copies RNA into cDNA, attaches the cell barcode and UMI to the cDNA. After this point, material from many droplets can be pooled without losing its address. ## Three kinds of barcode identify different units A **FASTQ file** stores sequences read by the sequencer together with their quality scores. In single-cell FASTQ data, `barcode` can refer to different addresses. A collection of molecules prepared for sequencing is a **library**, and a **sample index** distinguishes its source when several libraries are read together. | name | distinguishes | shared by | | --- | --- | --- | | sample index | libraries in a sequencing run | all molecules in one sample library | | cell barcode | droplets in one library | cDNA captured in one droplet | | UMI | original molecules in one cell | PCR reads from one captured molecule | A sample index separates sample A from sample B but cannot separate cells within sample A. The cell barcode separates cells, and the UMI separates molecules within a cell. ## Why expression counts UMIs instead of reads The small amount of cDNA from one cell must be amplified. One molecule may produce ten PCR copies while another produces two. Counting reads directly would make the first molecule appear more abundant. Reads with the same `cell barcode + gene + UMI` are therefore collapsed into one group. ```text CB=C1 gene=CD3D UMI=A7 read 1 CB=C1 gene=CD3D UMI=A7 read 2 -> PCR copy of the same molecule CB=C1 gene=CD3D UMI=K2 read 3 -> another original molecule result: CD3D count in cell C1 = 2 ``` UMIs are not perfect serial numbers. Two molecules can collide on the same UMI, and the UMI sequence can contain a sequencing error. Pipelines compare nearby UMI sequences and read support to correct likely errors. ## A 3′ count assay does not read the full transcript The 10x 3′ assay captures sequence near the poly(A) tail and efficiently counts molecules by gene across thousands of cells. In exchange, it does not cover the full transcript uniformly. This is well suited to gene-level expression and cell-type separation. It is more limited for comparing isoforms, different transcripts made from the same gene, or alternative splicing, changes in how exons are combined. Full-length methods such as Smart-seq2 are a better match for those questions. ## Summary Cell barcodes preserve the cell of origin, and UMIs identify RNA molecules before amplification. Collapsing PCR reads that share a cell, gene, and UMI reduces amplification bias in the count matrix. Next, [From FASTQ to a Cell by Gene Matrix](/en/lessons/single-cell-fastq-to-matrix/) follows these addresses through barcode validation, alignment, gene assignment, and sparse-matrix construction. --- ### Sources - 10x droplet barcoding: [Zheng et al., *Nature Communications* (2017)](https://www.nature.com/articles/ncomms14049) - UMI-based single-cell molecule counting: [Islam et al., *Nature Methods* (2014)](https://www.nature.com/articles/nmeth.2772) - Barcode and UMI processing: [10x Genomics Cell Ranger gene expression algorithm](https://www.10xgenomics.com/support/software/cell-ranger/latest/algorithms-overview/cr-gex-algorithm) - Full-length Smart-seq2: [Picelli et al., *Nature Protocols* (2014)](https://www.nature.com/articles/nprot.2014.006)