FR-HIT is an efficient fragment recruitment algorithm for next generation sequences against microbial reference genomes. It produces similar sensitivity of BLASTN, but runs at a 100 times higher speed. We applied FR-HIT, BLAST and several other alignment programs to recruit an Illumina dataset of 1 million 75-bp reads selected from a recent human gut microbiome study (Qin J, et al. Nature 2010, 464:59) to 194 public human gut bacterial reference genomes. BLAST recruited 475,584 reads and found 6,134,663 alignments in 241.5 hours. Mapping program BWA used 0.1 hours, but only produced 212,699 recruitments. FR-HIT recruited 523,868 reads and identified 5,780,580 alignments in 1.8 hours.
Current version: 0.7
The current version supports:
1. Multithreads parallel computing using OPENMP.
2. PSL output format.
3. Mask of low-complexity regions as lower cased sequences in reference database.
4. E-value cutoff.
Three perl scripts:
psl2sam.pl: transfer PSL format to SAM format.(This script is from samtools package)
frhit2pairend.pl: analysis FR-HIT output to get pair-end alignment information
binning-1.1.1: This program package performs taxonomy binning using output from FR-HIT. The algorithm is based on LCA (Lowest Common Ancestor).
Usage: fr-hit v0.7 [options]
-a <string> reads file, *.fasta format
-d <string> reference genome sequences file, *.fasta format
-o <string> output recruitments file
-e <double> e-value cutoff, default=10
-u <int> mask out repeats as lower cased sequence to prevent spurious hits? 1: yes; 0: no; default=1
-f <int> format control for output file,0:FR-HIT format; 1:PSL fromat, default=0
-k <int> k-mer size (8<=k<=12), default=11
-p <int> k-mer overlap of index (1<=p<-k), using small overlap for longer reads(454, Sanger), default=8
-c <int> sequence identity threshold(%), default=75
-g <int> use global or local alignment? 1:global; 0:local (need -m), default=0
-w <int> minimal read length to use 2bp k-mer index step to 454 long reads, default=1000
-m <int> minimal alignment coverage control for the read (g=0), default=30
-l <int> length of throw_away_reads, default=20
-t <int> maximum number of failed alingment attempts, default=20
-r [0,N] how to report alignment hits, 0:all; N:the best top N hits for one read, default=0
-n <int> do alignment for which chain? 0:both; 1:direct only; 2:complementary only. default=0
-b <int> band_width of alignment, default=4
-T [0,N] number of threads, default 1; with 0, all CPUs will be used
-h help
example:
./fr-hit -a 454reads-sample.fa -d 1000bacterialgenomes.fasta -c 90 -m 40 -w 120 -r 0 -o out.sop
The default output format of FR-HIT recruitment result file looks like:
ReadName ReadLength E-value AlignmentLength Begin End Strand Identity ReferenceSequenceName Begin End
1_lane2_1 75nt 8.3e-25 69 69 1 - 95.01% Acidaminococcus_D21 486841311 486841379
1_lane2_1 75nt 8.3e-25 69 69 1 - 95.23% Ruminococcus_5_1_39B_FAA 3450573 3450641
1_lane2_9 75nt 2.2e-25 64 1 64 + 98.81% Acidaminococcus_D21 7901322 7901385
1_lane2_9 75nt 2.2e-25 64 1 64 + 98.90% Alistipes_putredinis_DSM_17216 1029618 1029681
1_lane2_10 75nt 4.5e-23 72 1 72 + 93.32% Acidaminococcus_D21 453948881 453948952
1_lane2_10 75nt 4.5e-23 72 1 72 + 93.67% Prevotella_copri_DSM_18205 3128442 3128513
1_lane2_11 75nt 1.7e-22 74 75 2 - 91.21% Acidaminococcus_D21 451839012 451839085
1_lane2_11 75nt 1.7e-22 74 75 2 - 91.08% Prevotella_copri_DSM_18205 1018573 1018646
FR-HIT supports PSL output format and users can also use psl2sam.pl to convert PSL format to SAM format.
FR-HIT has no any dependency, so just clone FR-HIT repo, and build the fr-hit binary:
git clone https://github.com/Beifang/fr-hit.git
cd fr-hit
make
Now you can put the resulting binary where your $PATH can find it. If you have supermissions, then
I recommend dumping it in the system directory for locally compiled packages:
sudo mv fr-hit /usr/local/bin/