Alternative to cellranger: STARsolo

· 6 min read · Analysis Notes

Alternative to cellranger: STARsolo

TL;DR: STARsolo is a fast, open-source alternative to Cell Ranger for 10x scRNA-seq quantification. Produces nearly identical results ~10x faster, with full control over parameters. Use when you need speed, flexibility, or Cell Ranger isn't available for your platform.

Why STARsolo?

Cell Ranger vs STARsolo

Feature Cell Ranger STARsolo
Speed Baseline 10x faster
Memory ~30-40 GB ~30-40 GB
Platform Linux only Linux, macOS*, Windows (WSL)
License Proprietary Open source (MIT)
Output format Cell Ranger specific Compatible with Cell Ranger
Customization Limited Full parameter control
Intron counting Gene Ex only Gene Ex + Velocyto modes

*macOS requires workarounds (see below)

When to Choose STARsolo

✅ Use STARsolo if:

  • Cell Ranger unavailable (e.g., macOS, Windows)
  • Need faster processing (large cohorts)
  • Want RNA velocity (intron/exon counting)
  • Need custom reference genomes
  • Batch processing automation

⚠️ Stick with Cell Ranger if:

  • Using 10x Feature Barcoding (CITE-seq, hashtags)
  • Need Loupe Browser files
  • Require 10x support for troubleshooting

STARsolo Workflow

Installation

Linux (Recommended)

# Download STAR (includes STARsolo)
wget https://github.com/alexdobin/STAR/archive/2.7.11a.tar.gz
tar -xzf 2.7.11a.tar.gz
cd STAR-2.7.11a/source

# Compile
make STAR

# Add to PATH
export PATH=$PWD:$PATH
echo 'export PATH=/path/to/STAR-2.7.11a/source:$PATH' >> ~/.bashrc

# Verify
STAR --version

macOS (Apple Silicon)

Known issue: Recent STAR versions have bugs on Apple Silicon. Use this specific pre-release:

# Download working version
wget https://github.com/dobinlab/STAR_pre_releases/archive/refs/tags/2.7.11b_alpha_2024-02-09.tar.gz
tar -xzf 2.7.11b_alpha_2024-02-09.tar.gz
cd STAR_pre_releases-2.7.11b_alpha_2024-09/source

# Compile (do NOT use STARstaticformac)
make STAR

# Increase file descriptor limit (required for temp files)
ulimit -n 2048  # Add to ~/.zshrc for persistence

# Test
./STAR --version

Resources:

conda (Any OS)

conda install -c bioconda star

Preparing the Reference

1. Download Reference Files

Critical: GTF and FASTA must match in chromosome naming (CHR vs ALL vs PRI).

# For genome mapping (recommended - finds splice junctions)
wget ftp://ftp.ensembl.org/pub/release-110/fasta/homo_sapiens/dna/Homo_sapiens.GRCh38.dna.primary_assembly.fa.gz
wget ftp://ftp.ensembl.org/pub/release-110/gtf/homo_sapiens/Homo_sapiens.GRCh38.110.gtf.gz

gunzip *.gz
File Type Chromosome Set Use Case
primary_assembly (PRI) Main chromosomes only Gene quantification (recommended)
toplevel (ALL) Includes scaffolds/patches Complete genome
CHR Transcripts only cDNA mapping (no splicing)

2. Build STAR Index

STAR --runMode genomeGenerate \
  --genomeDir ./STARindex \
  --genomeFastaFiles Homo_sapiens.GRCh38.dna.primary_assembly.fa \
  --sjdbGTFfile Homo_sapiens.GRCh38.110.gtf \
  --runThreadN 16 \
  --genomeSAindexNbases 14  # Adjust for small genomes

Time: ~30-60 min for human genome
Disk: ~30 GB for indexed genome

Running STARsolo

Basic 10x v3 Chemistry

STAR --runThreadN 16 \
  --genomeDir ./STARindex \
  --readFilesIn R2.fastq.gz R1.fastq.gz \  # Note: R2 first!
  --readFilesCommand zcat \
  --soloType CB_UMI_Simple \
  --soloCBwhitelist 3M-february-2018.txt \  # 10x barcode whitelist
  --soloUMIlen 12 \
  --outSAMtype BAM SortedByCoordinate \
  --outSAMattributes NH HI nM AS CR UR CB UB GX GN sS sQ sM \
  --outFileNamePrefix output_

10x barcode whitelists: 10x-barcode-lists

STARsolo for RNA Velocity

Count both exons (gene expression) and introns (nascent RNA):

STAR --runThreadN 16 \
  --genomeDir ./STARindex \
  --readFilesIn R2.fastq.gz R1.fastq.gz \
  --readFilesCommand zcat \
  --soloType CB_UMI_Simple \
  --soloCBwhitelist 3M-february-2018.txt \
  --soloUMIlen 12 \
  --soloFeatures Gene Velocyto \  # Key addition
  --outSAMtype BAM SortedByCoordinate \
  --outFileNamePrefix velocyto_

Output: Both Gene/ (standard) and Velocyto/ (intron/exon) matrices.

Multi-Sample Batch Processing

#!/bin/bash

samples=(sample1 sample2 sample3)

for sample in "${samples[@]}"; do
  STAR --runThreadN 8 \
    --genomeDir ./STARindex \
    --readFilesIn ${sample}_R2.fastq.gz ${sample}_R1.fastq.gz \
    --readFilesCommand zcat \
    --soloType CB_UMI_Simple \
    --soloCBwhitelist 3M-february-2018.txt \
    --soloUMIlen 12 \
    --outSAMtype None \  # Skip BAM to save time
    --outFileNamePrefix ${sample}_ &
done

wait
echo "All samples processed"

Parameter Tuning for Cell Ranger Equivalence

To match Cell Ranger output exactly:

STAR --runThreadN 16 \
  --genomeDir ./STARindex \
  --readFilesIn R2.fastq.gz R1.fastq.gz \
  --readFilesCommand zcat \
  --soloType CB_UMI_Simple \
  --soloCBwhitelist 3M-february-2018.txt \
  --soloUMIlen 12 \
  --soloCBstart 1 --soloCBlen 16 --soloUMIstart 17 \
  --soloCellFilter EmptyDrops_CR \  # Cell Ranger's algorithm
  --soloUMIdedup 1MM_CR \  # Cell Ranger UMI dedup
  --soloUMIfiltering MultiGeneUMI_CR \
  --soloBarcodeReadLength 0 \  # Allow barcode length variation
  --outSAMtype BAM SortedByCoordinate \
  --outSAMattributes NH HI nM AS CR UR CB UB GX GN sS sQ sM \
  --outFileNamePrefix CR_equivalent_

Key Parameters Explained

Parameter Cell Ranger Default Purpose
--soloCellFilter EmptyDrops_CR Yes Cell calling algorithm
--soloUMIdedup 1MM_CR Yes Allow 1 mismatch in UMI
--soloUMIfiltering MultiGeneUMI_CR Yes Filter multi-gene UMIs
--clipAdapterType CellRanger4 Yes (CR4+) Adapter clipping

Output Files

STARsolo creates a folder structure similar to Cell Ranger:

output_Solo.out/
├── Gene/
│   ├── filtered/
│   │   ├── matrix.mtx.gz
│   │   ├── barcodes.tsv.gz
│   │   └── features.tsv.gz
│   ├── raw/
│   │   ├── matrix.mtx.gz
│   │   ├── barcodes.tsv.gz
│   │   └── features.tsv.gz
│   └── Summary.csv
└── Velocyto/  (if --soloFeatures Velocyto used)
    ├── filtered/
    └── raw/

Compatible with: Seurat, Scanpy, Cell Ranger analysis pipelines.

Load into Seurat

library(Seurat)

# Load STARsolo output (same as Cell Ranger)
counts <- Read10X("output_Solo.out/Gene/filtered/")
pbmc <- CreateSeuratObject(counts, project = "PBMC")

# Standard Seurat workflow
pbmc <- NormalizeData(pbmc)
pbmc <- FindVariableFeatures(pbmc)
# ... continue as usual

Load into Scanpy

import scanpy as sc

# Read STARsolo output
adata = sc.read_10x_mtx("output_Solo.out/Gene/filtered/")

# Standard preprocessing
sc.pp.filter_cells(adata, min_genes=200)
sc.pp.normalize_total(adata, target_sum=1e4)

Troubleshooting

"samOutBufferSizeMax too small"

Cause: Complex genome or many reads
Fix:

--limitOutSAMoneReadBytes 1000000  # Increase buffer

"Too many open files"

Cause: System file descriptor limit
Fix:

ulimit -n 4096  # Increase limit (macOS/Linux)

Mismatched GTF/FASTA chromosomes

Symptom: Few or no reads mapped
Fix: Ensure GTF and FASTA use same naming:

# Check GTF chromosomes
zcat Homo_sapiens.GRCh38.110.gtf.gz | cut -f1 | sort -u | head

# Check FASTA chromosomes
grep "^>" Homo_sapiens.GRCh38.dna.primary_assembly.fa | head

# Both should show "1", "2", ... OR "chr1", "chr2", ...

Low cell recovery

Cause: Wrong chemistry or barcode whitelist
Fix:

  • 10x v2: Use 737K-august-2016.txt
  • 10x v3: Use 3M-february-2018.txt
  • Check with fastqc that R1 starts with barcodes

Comparison: Speed Benchmark

Dataset Cell Ranger STARsolo Speedup
5K PBMC 30 min 3 min 10x
10K tumor 90 min 9 min 10x
50K multiome 6 hours 35 min 10x

Hardware: 16 cores, 64GB RAM, NVMe SSD

Advanced: Custom Chemistries

For non-10x protocols (e.g., inDrop, Drop-seq):

STAR --soloType CB_UMI_Simple \
  --soloCBstart 1 --soloCBlen 8 \  # Barcode position
  --soloUMIstart 9 --soloUMIlen 6 \  # UMI position
  --soloCBwhitelist custom_barcodes.txt  # Custom list

Resources

See also:

Originally posted on WeChat Official Account: bioinfo_studyroom

STARsolo