How to reduce STAR memory usage?

Optimize Mapping Parameters

• Reduce the genome index size by generating a genome index with fewer bases per bin. Use the `--genomeSAindexNbases` parameter with a lower value to decrease memory usage.

 

• Decrease the maximum number of mismatches allowed with the `--outFilterMismatchNmax` parameter to reduce computational requirements.

 

• Limit the number of alignments STAR retains by adjusting `--outFilterMultimapNmax` to a lower value, which can decrease memory usage.

Adjust Run-Time Parameters

• Set the `--limitBAMsortRAM` parameter to manage the maximum memory allocation for BAM sorting, thereby controlling STAR's memory use more effectively.

 

• Define the `--limitOutSJcollapsed` to restrict the total size of collapsed junctions. Modifying this parameter helps control memory consumption.

Use Efficient Data Management

• Downsample input data if possible, either by filtering out lower quality reads before alignment or by using a smaller subset for test runs.

 

• Employ the `--readFilesIn` option to directly read compressed files, which saves space and slightly reduces memory load during processing.

Utilize High-Performance Computing Resources

• Run STAR on a high-performance computing cluster with sufficient memory resources, specifying the memory allocation using `--genomeLoad` and setting it to parameters like `LoadAndKeep` or `NoSharedMemory` as appropriate.

 

• Prioritize nodes with larger RAM if available, ensuring that the system resources align with STAR's memory demands.

Tune Computational Threads

• Adjust the number of threads with the `--runThreadN` parameter to balance CPU load and memory usage, as more threads can increase memory requirements.

 

• Monitor and iteratively tune the number of threads based on the system's performance and available memory.