QC-Chain: fast and holistic quality control method for next-generation sequencing data

Abstract

Next-generation sequencing (NGS) technologies have been widely used in life sciences. However, several kinds of sequencing artifacts, including low-quality reads and contaminating reads, were found to be quite common in raw sequencing data, which compromise downstream analysis. Therefore, quality control (QC) is essential for raw NGS data. However, although a few NGS data quality control tools are publicly available, there are two limitations: First, the processing speed could not cope with the rapid increase of large data volume. Second, with respect to removing the contaminating reads, none of them could identify contaminating sources de novo, and they rely heavily on prior information of the contaminating species, which is usually not available in advance. Here we report QC-Chain, a fast, accurate and holistic NGS data quality-control method. The tool synergeticly comprised of user-friendly tools for (1) quality assessment and trimming of raw reads using Parallel-QC, a fast read processing tool; (2) identification, quantification and filtration of unknown contamination to get high-quality clean reads. It was optimized based on parallel computation, so the processing speed is significantly higher than other QC methods. Experiments on simulated and real NGS data have shown that reads with low sequencing quality could be identified and filtered. Possible contaminating sources could be identified and quantified de novo, accurately and quickly. Comparison between raw reads and processed reads also showed that subsequent analyses (genome assembly, gene prediction, gene annotation, etc.) results based on processed reads improved significantly in completeness and accuracy. As regard to processing speed, QC-Chain achieves 7-8 time speed-up based on parallel computation as compared to traditional methods. Therefore, QC-Chain is a fast and useful quality control tool for read quality process and de novo contamination filtration of NGS reads, which could significantly facilitate downstream analysis. QC-Chain is publicly available at: http://www.computationalbioenergy.org/qc-chain.html.

Figure 1. The overall workflow of QC-Chain.

Figure 2. Evaluation of read quality on real NGS data by QC-Chain using Parallel-QC.

(A) Summary of sequencing-quality evaluation. (B) Comparison of running time of Parallel-QC, FASTX_Toolkit and PRINSEQ. S1, S2: human saliva DNA samples; A1, A2: in-house sequenced algae DNA samples. All the sequences could be downloadable from http://computationalbioenergy.org/qc-chain.html.

Figure 3. Possible source species identified from simulated genomic data by rDNA-reads based method of QC-Chain.

(A) 18S reads distribution identified by rDNA-reads based method. (B) 16S reads distribution identified by rDNA-reads based method. (C) Quantitative distribution of the reads identified by random-reads based method.

Figure 4. Possible source species identified from simulated metagenomic data by rDNA-reads based method of QC-Chain.

(A) 18S reads distribution identified by rDNA-reads based method. (B) Quantitative distribution of the reads identified by random-reads based method.

Figure 5. Comparison of the results from clean, total and control reads of simulated metagenomic data.

(A) GC distribution pattern. (B) Rarefaction curve that could discriminate the richness of different bacterial species, Y-axies: species count, X-axis: the number of reads. (C) Functional categories based on COG database. Abundance of each category between the three datasets was compared pair-wise (*p<0.05; **p<0.01; NS: not significant).