Bassel Ghaddar 2025-10-15
Figure 1: Overview of PRISM.
PRISM identifies truly present microbial taxa in genomic sequencing data and eliminates falsely positive artifacts and contaminants. It identifies uniquely identifiable taxa using full-length mapping of a representative subsample of sequencing reads and taxa first identified with fast, k-mer-based taxonomic classification. It then employs a machine learning model to predict tissue-present microbes vs. contaminants based on multiple features engineered from read mapping and gene expression statistics (Fig. 1). PRISM is compatible with RNA-seq, WGS, 16S-seq, and scRNA-seq.
Please see the reference below for more information.
Please contact Bassel Ghaddar (bassel.ghaddar@gmail.com) for any questions.
PRISM requires the following dependencies:
- Kraken2: https://github.com/DerrickWood/kraken2
- Minimap2:https://github.com/lh3/minimap2
- STAR: https://github.com/alexdobin/STAR
- BLAST: https://ftp.ncbi.nlm.nih.gov/blast/executables/blast+/LATEST/
- Seqkit: https://bioinf.shenwei.me/seqkit/
- R packages: optparse, ShortRead, tidyverse, furrr, data.table, vegan
- Processed genbank files: https://drive.google.com/file/d/1nkdQ3GyG_FfQQI7nRWpzcQDJOTvULuGV/view?usp=sharing
After downloading this PRISM package and genbank folder, make sure to unzip the genbank folder and place it in the PRISM package directory.
There is a required BLAST database mapping file that must be created once:
One-time setup: create accession→taxid map for PRISM
-
Extract taxid-accession pairs from your BLAST database (edit path as needed)
blastdbcmd -db /path/to/blast/db/eg/core_nt \ -entry all -outfmt "%T %a" > /path/to/PRISM/taxid_accession_map.tsv -
Sort by taxid to create final map
sort -k1,1 /path/to/PRISM/taxid_accession_map.tsv > /path/to/PRISM/sorted_accession_map.txt
(Re-run only if BLAST database is updated)
PRISM is invoked with a single command-line R function PRISM.R:
PRISM.R
--samplesample name (e.g., ABC for ABC_1.fastq)--data_pathpath to FASTQ/FASTA files--kraken_pathpath to Kraken2 executable--kraken_db_pathpath to Kraken2 database--seqkit_pathpath to SeqKit executable--minimap2_pathpath to Minimap2 executable--minimap2_indexpath to Minimap2 host .mmi index--star_pathpath to STAR executable--star_genome_dirpath to STAR host genome index (for –genomeDir)--blast_pathpath to BLAST+ executables directory--blast_db_pathpath prefix to BLAST database (no extension)--prism_pathpath to PRISM repository (root folder)
--model_org_taxidspath to text file of model-organism taxids (default: prism_path/NA → bundled file)--pairedTRUE|FALSE (default TRUE)--fq1_endsuffix for read1 (default “_1.fastq”)--fq2_endsuffix for read2 (default “_2.fastq”)--barcode_onlyTRUE|FALSE; if TRUE, skip BLAST of read1 (default FALSE)--max_samplemaximum reads per taxon in subsample (default 100)--min_read_perminimum reads per X to analyze (default 1e4)--min_uniq_fracminimum unique k-mer ratio (default 5)--threadsnumber of threads for external tools (default 1)--min_qcovsminimum BLAST query coverage percent (default 80)--out_pathoutput directory (default: {data_path}/{sample}_prism/)--use_custom_dbTRUE|FALSE; build temporary subsetted BLAST DB (default TRUE)--custom_db_pathdirectory for temporary subsetted BLAST DB (default: {out_path}/data/customdb)--fasta_size_threshold_custom_dbminimum FASTA size (MB) to trigger custom DB (default 5)
PRISM creates a directory for each sample containing the main subfolder
data/ with all intermediate files and logs.
The key final outputs are:
X-results.csv— Final per-read table after all filtering and scoringX-counts.csv— Per-species summary of detected taxaX_1.fa/X_2.fa— Final PRISM FASTA(s) of retained microbial reads (headers annotated with taxid and accession)
This is the final BLAST alignment table after: 1. Identifying uniquely mappable microbial species, 2. Removing human, model organism, and vector sequences, 3. Resolving multi-mapping reads, and 4. Appending GenBank annotations and PRISM classification scores.
Each row corresponds to a single sequencing read.
Columns:
| Column | Description |
|---|---|
| id | Sequence identifier |
| read | Read number (1 or 2) for paired-end samples |
| staxids | NCBI taxon ID assigned |
| tax_name | Scientific name of the taxon |
| rank | Highest resolved phylogenetic rank (k,p,c,o,f,g,s) |
| sacc | BLAST subject accession |
| pos | Subject start position of alignment |
| qcovs | BLAST query coverage percent (0–100) |
| pident | Percent sequence identity |
| bitscore | BLAST bit score |
| gene | GenBank gene annotation (if available) |
| product | GenBank product annotation (if available) |
| pred | PRISM score (0 = likely contaminant, 1 = likely truly present) |
Aggregated species-level summary of reads and PRISM scores:
| Column | Description |
|---|---|
| tax_name | Scientific name |
| staxids | NCBI taxon ID |
| n | Number of reads assigned |
| pred | Mean PRISM score per taxon |
Final FASTA files containing only PRISM-retained reads.
Each header is annotated as: read_id | PRISM | staxids:{taxid}
sacc:{accession}
RNA-seq data from pancreatic cancer, using f942e6d6-f697-4141-8f1c-58933ca81751_1.fastq and f942e6d6-f697-4141-8f1c-58933ca81751_2.fastq from the CPTAC project.
Rscript \
/path/to/PRISM/PRISM.R \
--sample f942e6d6-f697-4141-8f1c-58933ca81751 \
--data_path /path/to/data/ \
--kraken_path /path/to/kraken2-master/kraken2 \
--kraken_db_path /path/to/kraken_db \
--seqkit_path /path/to/seqkit \
--minimap2_path /path/to/minimap2 \
--minimap2_index /path/to/host_hg38.mmi \
--star_path /path/to/STAR \
--star_genome_dir /path/to/star_index \
--model_org_taxids /path/to/PRISM/model_org_taxids.txt \
--blast_path /path/to/blast/2.16/bin/ \
--blast_db_path /path/to/BLAST/db \
--prism_path /path/to/PRISM/ \
--fq1_end _1.fastq \
--fq2_end _2.fastq \
--paired TThe files 0ac20066-3954-4a36-8ab3-c62a0c32d988-results.csv and
0ac20066-3954-4a36-8ab3-c62a0c32d988-counts.csv contain the main
results.
Species counts and PRISM scores:
library(tidyverse)
res = read.csv('./test data/0ac20066-3954-4a36-8ab3-c62a0c32d988-counts.csv')
tibble(res)## # A tibble: 13 × 4
## tax_name staxids n pred
## <chr> <int> <int> <dbl>
## 1 Enterococcus faecium 1352 5521 0.973
## 2 Escherichia coli 562 1531 0.105
## 3 Pseudomonas aeruginosa 287 277 0.009
## 4 Staphylococcus aureus 1280 262 0.005
## 5 Bacteroides fragilis 817 228 0.617
## 6 Cupriavidus taiwanensis 164546 199 0.002
## 7 Paucibacter sediminis 3019553 175 0
## 8 Pseudomonas tolaasii 29442 161 0.002
## 9 Cutibacterium acnes 1747 120 0.006
## 10 Pseudomonas yamanorum 515393 99 0.004
## 11 Parvimonas micra 33033 43 0.006
## 12 Prevotella nigrescens 28133 43 0.002
## 13 Malassezia restricta 76775 24 0.001
Examining read-level microbial gene/product data:
library(tidyverse)
res = read.csv('./test data/0ac20066-3954-4a36-8ab3-c62a0c32d988-results.csv')
tibble(res)## # A tibble: 8,683 × 13
## id read staxids tax_name rank sacc gene product pos qcovs pident
## <chr> <int> <int> <chr> <chr> <chr> <chr> <chr> <int> <int> <dbl>
## 1 K00270:… 2 562 Escheri… S S429… "" "" 105 100 100
## 2 K00270:… 2 1352 Enteroc… S MK33… "" "16S r… 624 100 100
## 3 K00270:… 1 1352 Enteroc… S LC56… "" "" 124 99 90.8
## 4 K00270:… 1 1352 Enteroc… S LC56… "" "" 124 100 92.2
## 5 K00270:… 1 1352 Enteroc… S CP13… "" "" 263 99 100
## 6 K00270:… 1 3019553 Pauciba… S CP11… "" "23S r… 899421 100 100
## 7 K00270:… 2 562 Escheri… S AP02… "" "23S r… 457497 100 100
## 8 K00270:… 1 3019553 Pauciba… S CP11… "" "23S r… 899421 100 100
## 9 K00270:… 2 562 Escheri… S CP09… "" "23S r… 468371 100 100
## 10 K00270:… 1 1352 Enteroc… S LC56… "" "" 124 100 89.6
## # ℹ 8,673 more rows
## # ℹ 2 more variables: bitscore <dbl>, pred <dbl>
Ghaddar B, Blaser M, De S. Revisiting the cancer microbiome using PRISM. bioRxiv 2025