5x throughput improvement

README.md

@@ -52,15 +52,17 @@ See below for information on usage and local installation.
     ```
     See full options below:
     ```
-    usage: pangolin [-h] [-c COLUMN_IDS] [-m {False,True}] [-s SCORE_CUTOFF] [-d DISTANCE] variant_file reference_file annotation_file output_file
+    usage: pangolin [-h] [-c COLUMN_IDS] [-m {False,True}] [-s SCORE_CUTOFF] [-d DISTANCE] [--score_exons {False,True}] [--loglevel {DEBUG,INFO,WARNING,ERROR,CRITICAL}] [--tmpdir TMPDIR] [--variant_batchsize VARIANT_BATCHSIZE]
+                [--tensor_batchsize TENSOR_BATCHSIZE]
+                variant_file reference_file annotation_file output_file
 
     positional arguments:
       variant_file          VCF or CSV file with a header (see COLUMN_IDS option).
       reference_file        FASTA file containing a reference genome sequence.
       annotation_file       gffutils database file. Can be generated using create_db.py.
       output_file           Prefix for output file. Will be a VCF/CSV if variant_file is VCF/CSV.
-
-    optional arguments:
+    
+    options:
       -h, --help            show this help message and exit
       -c COLUMN_IDS, --column_ids COLUMN_IDS
                             (If variant_file is a CSV) Column IDs for: chromosome, variant position, reference bases, and alternative bases. Separate IDs by commas. (Default: CHROM,POS,REF,ALT)
@@ -70,6 +72,15 @@ See below for information on usage and local installation.
                             Output all sites with absolute predicted change in score >= cutoff, instead of only the maximum loss/gain sites.
       -d DISTANCE, --distance DISTANCE
                             Number of bases on either side of the variant for which splice scores should be calculated. (Default: 50)
+      --score_exons {False,True}
+                            Output changes in score for both splice sites of annotated exons, as long as one splice site is within the considered range (specified by -d). Output will be: gene|site1_pos:score|site2_pos:score|...
+      --loglevel {DEBUG,INFO,WARNING,ERROR,CRITICAL}
+                            Set the logging level. (Default: INFO)
+      --tmpdir TMPDIR       Location to create temporary directory for storing intermediate files.
+      --variant_batchsize VARIANT_BATCHSIZE
+                            Number of variants to score in a single CPU batch. (Default: 1280)
+      --tensor_batchsize TENSOR_BATCHSIZE
+                            Number of variants to process in a single GPU batch. (Default: 128)
     ```
 
 ### Usage (custom)

docker/Dockerfile

@@ -0,0 +1,57 @@
+######################################
+## CONTAINER FOR GPU based pangolin ##
+######################################
+
+# start from the cuda docker base
+from nvidia/cuda:12.0.0-runtime-ubuntu22.04
+
+## needed apt packages
+ARG BUILD_PACKAGES="wget git bzip2"
+# needed conda packages 
+ARG CONDA_PACKAGES="python==3.10.8 pip==25.0 pandas==2.2.3 pyfastx==0.8.4 gffutils==0.13 pytorch==2.5.1 torchvision==0.20.1 torchaudio==2.5.1 pytorch-cuda==12.4 pysam==0.20.0"
+ARG CONDA_CHANNEL="-c nvidia -c pytorch -c conda-forge -c anaconda -c bioconda"
+##  ENV SETTINGS during runtime
+ENV LANG=C.UTF-8 LC_ALL=C.UTF-8
+ENV PATH=/opt/conda/bin:$PATH
+ENV DEBIAN_FRONTEND noninteractive
+
+## AUTHOR
+ENV AUTHOR="Geert Vandeweyer"
+ENV EMAIL="geert.vandeweyer@uza.be"
+
+# For micromamba:
+SHELL ["/bin/bash", "-l", "-c"]
+ENV MAMBA_ROOT_PREFIX=/opt/conda/
+ENV PATH=/opt/micromamba/bin:/opt/conda/bin:$PATH
+
+
+## INSTALL 
+RUN apt-get -y update  && \
+    apt-get -y install $BUILD_PACKAGES && \
+    apt-get clean && \
+    rm -rf /var/lib/apt/lists/*
+
+
+# conda packages
+RUN mkdir /opt/conda && \
+    mkdir /opt/micromamba && \
+    wget -qO - https://micromamba.snakepit.net/api/micromamba/linux-64/0.23.0 | tar -xvj -C /opt/micromamba bin/micromamba && \
+    # initialize bash
+    micromamba shell init --shell=bash --prefix=/opt/conda && \
+    # remove a statement from bashrc that prevents initialization
+    grep -v '[ -z "\$PS1" ] && return' /root/.bashrc  > /opt/micromamba/bashrc && \
+    mv /opt/micromamba/bashrc /root/.bashrc && \
+    source ~/.bashrc && \
+    # activate & install base conda packag
+    micromamba activate && \
+    micromamba install -y $CONDA_CHANNEL $CONDA_PACKAGES && \
+    micromamba clean --all --yes
+
+# my fork of pangolin : has gpu optimizations
+RUN cd /opt/ && \
+    git clone https://github.com/geertvandeweyer/pangolin.git && \
+    cd pangolin && \
+    pip install .  
+
+# ADD annotation data
+CMD ["pangolin", "--help"]