Skip to content

3. Cap Prediction

Capfinder provides a command-line interface to predict RNA cap types using BAM and POD5 files. Here's how to use the predict-cap-types function:

Usage

capfinder predict-cap-types [OPTIONS]

Description

This command predicts RNA cap types using BAM and POD5 files.

Required Options

  • --bam_filepath or -b: Path to the BAM file generated using the preprocessing step

  • --pod5_dir or -p: Path to directory containing POD5 files

  • --output_dir or -o: Path to the output directory for prediction results and logs

Additional Options

  • --n_cpus or -n: Number of CPUs to use for parallel processing. Default is 1
Multiple CPUs are used during processing for POD5 file and BAM data (Step 1/5). Increasing this number speeds up POD5 and BAM processing. For inference (Step 4/5), only a single CPU is used no matter how many CPUs you have specified. For faster inference, have a GPU available (it will be detected automatically) and set dtype to float16
  • --dtype or -d: Data type for model input. Valid values are float16, float32, or float64. Default is float16
Without a GPU, use float32 or float64 for better performance. If you have a GPU, then use float16 for faster inference
  • --batch_size or -bs: Batch size for model inference. Default is 128
Larger batch sizes can speed up inference but require more memory. If the code crashes during step 4/5, you have probably set too high a batch size.

  • --plot-signal / --no-plot-signal: Whether to plot extracted cap signal or not. Default is --no-plot-signal

  • --custom_model_path or -m: Path to a custom model (.keras) file. If not provided, the default pre-packaged model will be used.

Saving plots can help you plot the read's signal, and plot the signal for cap and flanking bases(±5).
  • --debug / --no-debug: Enable debug mode for more detailed logging. Default is --no-debug
The option can prints which function is creating a particular log output. This is helpful during code debugging.
  • --refresh-cache / --no-refresh-cache: Refresh the cache for intermediate results. Default is --no-refresh-cache
If you input data has changed (for example you added one more POD5 file in your POD5 directory) then you must use --refresh-cache to compute all steps again and not load them from cache that hold results from your previous run.
  • --help: Show the help message and exit

Example

capfinder predict-cap-types \
    --bam_filepath /path/to/sorted.bam \
    --pod5_dir /path/to/pod5_dir \
    --output_dir /path/to/output_dir \
    --n_cpus 100 \
    --dtype float16 \
    --batch_size 256 \
    --no_plot_signal \
    --no-debug \
    --no-refresh-cache

Tips

  1. CPU Usage:

    • Increase --n_cpus for faster processing of POD5 and BAM data
    • CPU count doesn't affect inference speed (Step 4/5)

  2. GPU Acceleration:

    • If you have a GPU, use --dtype float16 for faster inference
    • Without a GPU, float32 or float64 may perform better

  3. Batch Size:

    • Larger batch sizes can speed up inference but require more memory
    • Adjust --batch_size based on your system's capabilities

  4. Plotting:

    • Use --no-plot-signal to skip signal plotting for faster processing

  5. Debugging:

    • Enable --debug for detailed logging when troubleshooting

  6. Caching:

    • Use --refresh-cache if you've made changes to input data and need to regenerate intermediate results

For more detailed information, run capfinder predict-cap-types --help.