How Does Ben Callahan’s DADA2 Improve the Accuracy and Resolution of Microbiome Analysis?

TL;DR

Ben Callahan’s DADA2 improves microbiome analysis by modeling amplification and sequencing errors, then inferring the exact sample sequences and abundances behind the reads. Unlike OTU analysis, which collapses sequences within a three percent radius and loses information, DADA2 can distinguish errors from biological variation, including rare variants. Read on to understand the amplicon inference problem and DADA2’s alternative approach.

Transcript

thanks for having me come and talk I feel like since my advisor is here i should mention i'm now in the stats department so this is sort of the microbial appendix to today's today's talk so i'm going to be talking again about microbes and particularly it's a it's a method talk before I jump into the method I just wanted to take a moment look here a... Read More

Key Insights

• 😃 Microbes make up a significant portion of life on Earth, with the majority being microbial and invisible to the naked eye.
• ❓ Amplicon sequencing has revolutionized the study of microbial populations and the human microbiome.
• 🎁 The amplicon inference problem presents challenges in accurately inferring sample sequences and their abundances due to errors introduced during amplification and sequencing.
• 👻 The DADA2 method offers an alternative approach to the OTU method, allowing for the inference of sample sequences and their abundances with better accuracy.
• 👻 The DADA2 method can effectively differentiate between errors and biological variations, allowing for the resolution of rare variants.
• 🉐 By using the DADA2 method, researchers can gain a more accurate understanding of microbial communities, such as in the study of the vaginal microbiome during pregnancy.
• 🫢 The DADA2 method is available as an open-source R package, allowing other researchers to use and explore this method.

Questions & Answers

Q: How does Ben Callahan’s DADA2 improve microbiome analysis?

DADA2 treats amplicon analysis as an unsupervised learning problem and attempts to infer the exact sample sequences and their abundances. By modeling amplification and sequencing errors, it can distinguish those errors from biological variation and resolve rare variants more accurately.

Q: What is amplicon sequencing?

Amplicon sequencing uses PCR with a primer specific to a particular bacterial gene, then sequences that amplified piece of DNA deeply. It does not reveal the whole genome, but the selected barcode can be used to survey the bacterial community.

Q: What is the amplicon inference problem?

The amplicon inference problem is determining the true sample sequences and their abundances from a stack of sequencing reads. The challenge is that PCR amplification and sequencing introduce errors, adding noise to the observed reads.

Q: Why is amplicon sequence inference difficult for rare variants?

A read that differs from earlier reads may represent either an error or a genuine rare variant. Even the 10,000th read could be a new community variant, so greater sequencing depth does not automatically resolve the uncertainty.

Q: What is an operational taxonomic unit, or OTU?

An operational taxonomic unit is created by taking an abundant sequence, drawing a three percent radius around it, and collapsing the sequences within that radius together. This reduces the effect of errors but also discards sequence information.

Q: How does DADA2 differ from the OTU method?

The OTU method groups similar sequences within a fixed radius, while DADA2 tries to infer the exact sequences and abundances that generated the reads. DADA2 therefore avoids the information loss caused by coarse OTU clustering.

Q: How does DADA2 model sequencing errors?

DADA2 aligns a sequencing read with a proposed sample sequence and evaluates matches and mismatches across the sequence. It multiplies position-specific probabilities while treating each nucleotide independently.

Q: What factors affect DADA2’s nucleotide error probabilities?

The probability at each nucleotide can depend on the true nucleotide, the read nucleotide, and the quality score at that position. The model also includes a batch effect because sequencing runs differ.

Summary & Key Takeaways

• Definition: Amplicon sequencing amplifies a selected bacterial gene with PCR and deeply sequences that DNA barcode.

• Number: More than 10,000 studies using amplicon sequencing were published since 2010.

• Definition: The amplicon inference problem seeks the true sample sequences and abundances underlying noisy sequencing reads.

• Definition: PCR amplification and sequencing can both introduce errors into amplicon reads.

• Definition: An OTU is formed by collapsing similar sequences around an abundant sequence.

• Number: The OTU method described uses a three percent sequence radius.

• Compare: OTU clustering reduces error impact but loses information through coarse grouping.

• Compare: DADA2 infers exact sample sequences and abundances instead of constructing OTUs.

• Step 1: Align a sequencing read with a proposed sample sequence and identify nucleotide matches and mismatches.

• Step 2: Multiply the position-specific probabilities while treating each nucleotide independently.

• Tool: DADA2 models probabilities using true nucleotide, read nucleotide, quality score, and sequencing-run batch effects.

• Tool: DADA2 is available as an open-source R package for microbiome researchers.