GSE71536 Processing Pipeline

Publication

Nature genetics (2017) — PMID 28092684

Dataset

Biallelic mutations in the 3’ exonuclease TOE1 cause Pontocerebellar Hypoplasia Type 7 and result in snRNA processing defects

1. 1

   Demultiplexing: Reads were demultiplexed based on the presence of the gene-specific primer at the beginning of R2, and a random-mer was extracted off of the beginning of R1.

   eclip_primer_umi_demultiplex.py v(Inferred with models/gemini-2.5-flash) GitHub

   $ Bash example

   # This script performs demultiplexing for eCLIP data.
   # It identifies reads based on a gene-specific primer at the 5' end of R2,
   # and extracts a random-mer (UMI) from the 5' end of R1, appending it to the read ID.
   
   # Define input files (placeholders)
   READ1_INPUT="sample_R1.fastq.gz"
   READ2_INPUT="sample_R2.fastq.gz"
   
   # Define parameters inferred from the description
   # Replace with actual primer sequence and random-mer length
   GENE_SPECIFIC_PRIMER_SEQ="YOUR_GENE_SPECIFIC_PRIMER_SEQUENCE" # Example: "GATCGGAAGAGCACACGTCTGAACTCCAGTCAC"
   RANDOM_MER_LENGTH=10 # Example: 10 bp random-mer
   
   # Define output files
   OUTPUT_PREFIX="demultiplexed_sample"
   READ1_OUTPUT="${OUTPUT_PREFIX}_R1.fastq.gz"
   READ2_OUTPUT="${OUTPUT_PREFIX}_R2.fastq.gz"
   LOG_FILE="${OUTPUT_PREFIX}.log"
   
   # Installation (example for tools that might be used internally by such a script)
   # # conda install -c bioconda umi_tools cutadapt
   
   # Execute the inferred custom demultiplexing script.
   # This script is assumed to be a custom Python script (e.g., from the eCLIP pipeline)
   # that combines primer detection/filtering in R2 and random-mer extraction in R1.
   # It would typically filter out reads not containing the primer and move the random-mer
   # from R1 to the read header of both R1 and R2.
   python eclip_primer_umi_demultiplex.py \
       --read1 "${READ1_INPUT}" \
       --read2 "${READ2_INPUT}" \
       --primer_r2 "${GENE_SPECIFIC_PRIMER_SEQ}" \
       --random_mer_length_r1 "${RANDOM_MER_LENGTH}" \
       --output_prefix "${OUTPUT_PREFIX}" \
       --log_file "${LOG_FILE}"

   Copy View on GitHub
2. 2

   Read trimming: Reads were trimmed at the 3' end using the cutadapt program, with options "--discard-untrimmed flag -a Gene_Specific_Primer"

   cutadapt v3.4 (Inferred) GitHub

   $ Bash example

   # Install cutadapt (example using conda)
   # conda install -c bioconda cutadapt=3.4
   
   # Trim reads at the 3' end using the specified primer, discarding reads that do not contain the primer.
   cutadapt --discard-untrimmed -a "Gene_Specific_Primer" -o trimmed_reads.fastq.gz raw_reads.fastq.gz

   Copy View on GitHub
3. 3

   Tail counting: Non-PCR duplicate reads were counted by summing reads with the same sequence that did not have the same randommer sequence as any other read.

   umi_tools (Inferred with models/gemini-2.5-flash) v1.1.2 GitHub

   $ Bash example

   # Install umi_tools if not already installed
   # conda install -c bioconda umi_tools
   
   # Input BAM file (e.g., aligned reads with UMIs in read IDs)
   INPUT_BAM="aligned_reads_with_umis.bam"
   # Output deduplicated BAM file
   OUTPUT_DEDUP_BAM="deduplicated_reads.bam"
   
   # Perform UMI deduplication. This step identifies and removes PCR duplicates
   # based on mapping position and Unique Molecular Identifier (UMI/randommer).
   # Reads with the same genomic coordinates but different UMIs are considered unique.
   # The description "summing reads with the same sequence that did not have the same randommer sequence as any other read"
   # directly corresponds to the output of this deduplication process.
   umi_tools dedup \
       -I "${INPUT_BAM}" \
       -S "${OUTPUT_DEDUP_BAM}" \
       --extract-umi-method=read_id \
       --umi-separator=":" \
       --method=unique \
       --log="umi_tools_dedup.log"
   
   # To perform the "tail counting" (i.e., counting the non-PCR duplicate reads) after deduplication:
   # samtools view -c "${OUTPUT_DEDUP_BAM}"

   Copy View on GitHub
4. 4

   Counts were performed for unique only (hamming distance > 0), as well as randommers at least 1 nt apart (hamming distance >1)

   umi_tools (Inferred with models/gemini-2.5-flash) v1.1.2 GitHub

   $ Bash example

   # Install umi_tools
   # conda install -c bioconda umi_tools=1.1.2
   
   # Example usage:
   # This command assumes that UMIs have already been extracted from reads
   # and appended to the read names (e.g., using `umi_tools extract`).
   # The input BAM file should be sorted by coordinate or read name, depending on the UMI extraction method.
   
   # Input BAM file with UMIs in read names (e.g., from a previous alignment and UMI extraction step)
   INPUT_BAM="aligned_reads_with_umis.bam"
   # Output BAM file containing deduplicated reads, with one read per unique UMI group
   OUTPUT_BAM="deduplicated_reads.bam"
   # Log file for umi_tools output and statistics
   LOG_FILE="umi_tools_dedup.log"
   
   # Perform UMI deduplication based on sequence similarity, correcting for sequencing errors.
   # --method=cluster: This method groups UMIs based on their sequence similarity, allowing for error correction.
   # --s_dist=1: This parameter specifies the maximum Hamming distance between two UMIs to be considered the same UMI.
   #             Setting s_dist=1 means that UMIs differing by 0 or 1 nucleotide are grouped together.
   #             This addresses the description's criteria:
   #             - "Counts were performed for unique only (hamming distance > 0)": The output BAM will contain reads with unique UMIs
   #               after deduplication, where 'unique' implies distinct UMI sequences or distinct UMI groups.
   #             - "as well as randommers at least 1 nt apart (hamming distance >1)": By grouping UMIs with s_dist=1,
   #               UMIs that are 0 or 1 mismatch apart are considered part of the same 'randommer' group. This implies that
   #               only 'randommers' (i.e., the representative UMIs of the clusters) that are separated by a Hamming distance
   #               of 2 or more (i.e., >1) are counted as truly distinct, fulfilling the specified criterion.
   umi_tools dedup \
       --input="${INPUT_BAM}" \
       --output="${OUTPUT_BAM}" \
       --method=cluster \
       --s_dist=1 \
       --log="${LOG_FILE}"

   Copy View on GitHub