Generating consensus sequence from bam file

Question

[chparada]

Hi,

I am trying to generate consensus sequence from a bam file obtained after mapping SRA reads to a reference genome.

I used the following commands:

bwa mem ref.fasta SRR_1.fastq SRR_2.fastq > bwa.sam
samtools view -b -F 4 bwa.sam > bwa_aligned.bam
samtools index bwa_aligned.bam

I am not sure how to generate the consensus sequence that I have in mind. In case I don't explain this well. I made a diagram:

===========================================================>ref.fasta
- -- ---- ----      ----- --- --- -      --    ------ ----- 
------ --- ---     -------- --- --       ---- -      --- -->SRR_reads_mapping



==============  +  ================  +   ==================> consensus_sequence.fasta

Please let me know if you have any advice on this.

Cheers!!!

[Asaf]

What do you mean by consensus sequence? The most frequent nucleotide in each position? The variants? There are plenty of tools that can give you the reading in each position, try bcftools mpileup for instance.

[chparada]

Thank you for your answers.

I meant obtaining the most frequent nucleotide in each position. From the mapped reads, I want to been able to obtain a consensus sequence in single fasta file. I am going to try bcftools mpileup.

[vsingh]

see below my answer thanks

Answer 1

[vsingh]

1) Map short reads against reference gene sequence

# Create bowtie2 database
bowtie2-build REFERENCE.fasta REF_DB

# bowtie2 mapping
bowtie2 -x REF_DB -U SAMPLE.fastq --no-unal -S SAMPLE.sam

# samtools:  sort .sam file and convert to .bam file
samtools view -bS SAMPLE.sam | samtools sort - -o SAMPLE.bam

2) Get consensus sequence from .bam file

# Get consensus fastq file
samtools mpileup -uf REFERENCE.fasta SAMPLE.bam | bcftools call -c | vcfutils.pl vcf2fq > SAMPLE_cns.fastq

  # vcfutils.pl is part of bcftools

# Convert .fastq to .fasta and set bases of quality lower than 20 to N
seqtk seq -aQ64 -q20 -n N SAMPLE_cns.fastq > SAMPLE_cns.fasta

Answer 2

[aswinssoman]

The lastest version of samtools_V1.16 allows you to generate a consensus from a SAM, BAM or CRAM file based on the contents of the alignment records. The consensus is written either as FASTA, FASTQ, or a pileup oriented format. This is selected using the -f FORMAT option.

Eg:

samtools consensus -f fasta in.bam -o cons.fa 

Answer 3

[nsmi8446]

You could call variants (using whatever variant calling software you like, GATK, freebayes etc.) from your .bam file and then use vcf-consensus (http://vcftools.sourceforge.net/perl_module.html#vcf-consensus) to build your consensus sequence. The code below should work:

cat ref.fa | vcf-consensus file.vcf.gz > out.fa

[cfos4698]

I agree with nsmi8446. This is a nice concise way to solve your problem.

Answer 4

[waldeyr]

samtools mpileup -uf my_reference.fna my_file.bam | bcftools view -cg - | vcfutils.pl vcf2fq > my_consensus.fq

Answer 5

[trausch]

Alfred has a consensus mode that extracts all reads at a given alignment position and then runs a multiple sequence alignment computation with consensus generation. It's primarily for long reads but I think it also works for short reads.

alfred consensus -t ill -f bam -p chr4:500500 input.bam

Answer 6

[chen]

try gencore: https://github.com/OpenGene/gencore

Generate consensus reads to reduce sequencing noises and remove duplications

Answer 7

[guillaume.rbt]

you can use GATK FastaAlternateReferenceMaker to generate the consensus sequence based on a SNP calling : https://software.broadinstitute.org/gatk/documentation/tooldocs/3.8-0/org_broadinstitute_gatk_tools_walkers_fasta_FastaAlternateReferenceMaker.php

Answer 8

[DareDevil]

samtools mpileup -uf reference.fasta sorted_aligned_reads.bam | bcftools call -c | vcfutils.pl vcf2fq > consensus.fastq

Here's a breakdown of the command:

samtools mpileup: Generates a pileup of aligned reads at each position in the reference genome.

-u: Output in uncompressed BAM format.

-f reference.fasta: Specifies the reference genome in FASTA format.

sorted_aligned_reads.bam: The sorted and indexed BAM file.

bcftools call: Calls the consensus genotype for each position based on the pileup.

-c: Output the consensus genotype.

vcfutils.pl vcf2fq: Converts the consensus genotype in VCF format to FASTQ format, representing the consensus sequence.

consensus.fastq: The output file containing the consensus sequence in FASTQ format.

Please make sure to replace reference.fasta with the filename of your reference genome and sorted_aligned_reads.bam with the appropriate name of your sorted and indexed BAM file.

After running this script, you should obtain the consensus sequence in the consensus.fastq file. If you prefer a FASTA format instead of FASTQ, you can use tools like seqtk or fastq_to_fasta to convert the FASTQ file to FASTA format if needed.

[singhankit973]

samtools mpileup is depriciated, so replace it with bcftools mpileup

Answer 9

[jkbonfield]

Since this was written, samtools now has a samtools consensus command. See the man page for controlling the coordinate system and whether indels are there.

It's much faster than the bcftools route, but potentially not as accurate in some scenarios (and more accurate in others, particularly with long read data). Something to try anyway.

[cfos4698]

This is interesting, thank you. Do you have any further information on why it might be more accurate with long read data? And what roughly constitutes 'long' in this case? For example, we have reads derived from amplicon sequencing using ONT that are on average 1.2 kb in length, would you expect a possible improvement with samtools consensus compared to the other approach of SNP calling --> bcftools consensus?

[jkbonfield]

A late reply I know. The main difference is the error models associated with long-read technologies - the actual read length doesn't matter really here. Consensus generation assuming most errors are substitution ones isn't so realiable when many newer technologies have indel as the primary error instead.