Tool: I built BioMark — like Geekbench but for bioinformatics. Tells you honestly if your laptop can run BWA-MEM2, STAR, Cell Ranger and more. No tools needed.
BioMark simulates 8 real bioinformatics pipelines on your machine without requiring any tools to be installed. It gives you honest PASS/FAIL per pipeline step and generates an HTML report with scores and hardware recommendations.
What it benchmarks:
- DNA/WES (BWA-MEM2, GATK, ANNOVAR)
- Bulk RNA-seq (STAR, DESeq2, clusterProfiler)
- scRNA-seq (Cell Ranger, Seurat, Monocle3)
- Protein Structure (AlphaFold2, ESMFold)
- Epigenomics (ChIP-seq, ATAC-seq)
- Metagenomics (Kraken2, QIIME2)
- Long Read (Nanopore, PacBio)
- Genome Assembly (SPAdes, Flye, Hifiasm)
Quick start:
git clone https://github.com/shahbazigenomics/BioMark.git
cd BioMark
pip install -r requirements.txt
python src/main.py --all
Example result on MacBook Air M1 8GB:
- DNA/WES: 15/100 — RAM+SSD insufficient
- RNA-seq: 20/100 — STAR needs 32GB RAM
- Assembly: 0/100 — All assemblers fail
- Overall: 13/100
Honest. No sugarcoating. That's the point.
GitHub: https://github.com/shahbazigenomics/BioMark
Would love to hear your scores and feedback!
• 306 views
•
link
updated
by
Istvan Albert
•
written
by
shahbazigenomics •
•
written
by
shahbazigenomics •
0 answers
No answers yet.
Log in to answer this question.
Interesting idea, I see you actually run some tools using simulated data to evaluate performance.
I hope this makes easier to show why people should not analyze data in their laptops.
Very Interesting!
Interesting project. It would be cool if you could also benchmark structural variant calling pipelines in relation to G4-rich regions. Some tools fail not just because of RAM, but because of the complexity of the secondary structure they are trying to align.
With laptops the main issue is heat dissipation and the subsequent CPU throttling to avoid overheating. Also I want to mention that one can easily run HiSat2 or Salmon on a Macbook Air and with that quickly perform an RNA-Seq analysis. Giving it a 20/100 is not realistic.
the benchmark should operate on task type and genome size, rather than task alone.
I ran a number of benchmarks myself, and based on my observations, even a MacBook Air can handle many genomics workloads on genomes up to about 50-100 Mb in size, including variant calling and RNA-seq analysis. Performance issues begin to appear around the 100 Mb mark, with the main limitation being thermal throttling. Because the system cannot dissipate heat effectively under sustained load, CPU performance drops substantially, leading to slower execution times.
The Mac M architecture is a technological marvel.
100% genome size needs to be considered. Also many of those tools have parameters that can change RAM use...