Re: Hyper Electronics Mappers Keygen Software
Adalrico Drury
DNA methylation plays a crucial role in higher organisms. Coupling bisulfite treatment with next generation sequencing enables the interrogation of 5-methylcytosine sites in the genome. However, bisulfite conversion introduces mismatches between the reads and the reference genome, which makes mapping of Illumina and SOLiD reads slow and inaccurate. BatMeth is an algorithm that integrates novel Mismatch Counting, List Filtering, Mismatch Stage Filtering and Fast Mapping onto Two Indexes components to improve unique mapping rate, speed and precision. Experimental results show that BatMeth is faster and more accurate than existing tools. BatMeth is freely available at
DNA methylation modifies the nucleotide cytosine by the addition of methyl groups to its C5 carbon residue by DNA methyltransferases [1]. This modification can be inherited through cell division and it plays an important role in many biological processes, such as heterochromatin and transcriptional silencing [2, 3], imprinting genes [4], inactivating the chromosome [5] and silencing of repetitive DNA components in healthy and diseased (including cancerous) cells [6, 7]. Methylation analysis can also be used to diagnose pre-natal Down's syndrome [8]. Thus, the genome-wide methylation profiles of different tissues are important to understand the complex nature and effects of DNA methylation.
One important step in calling methylation of a genome is to map bisulfite reads. Mapping of bisulfite reads is different from that of ChIP-Seq and RNA-Seq data since the non-methylated Cs are converted to Ts by bisulfite treatment and subsequent PCR. The bisulfite reads are difficult to map to the reference genome due to the high number of mismatches between the converted Ts and the original Cs. For mapping Illumina bisulfite reads, the pioneering published methods are BSMAP [15] and RMAP [16]. BSMAP aligns a bisulfite read to the reference genome by first enumerating all C-to-T combinations within a user-defined length k seed of the reads; then, through hashing, BSMAP aligns the seeds onto the genome and putative alignments are extended and validated with the original reads. After this step, BSMAP can output an unambiguous hit for each read, if available. BRAT [17] uses a similar strategy as BSMAP. It converts the reference genome into a TA reference and a CG reference (each converted reference uses one bit per base). Using a 36-mer hash table, BRAT aligns the first 36 bases of every read and its 1-neighbors on the two converted references to identify possible alignments. RMAP uses layered seeds as a bit-mask to select a subset of the bases in the reads and constructs a hash table to index all the reads. However, these seed-hash-based approaches are slow.
Subsequently, several methods were proposed to map bisulfite reads onto the converted genomes. MethylCoder [18] surfaced as a bisulfite read mapper that uses GSNAP [19] to do a primary mapping of in silico converted reads (that is, all Cs in the reads are converted to Ts) onto a converted reference genome (that is, all Cs in the genome are converted to Ts). Those reads that fail to map onto the converted genome will be remapped again in their original forms onto the original reference. BS-Seeker [20] and Bismark [21] use a similar conversion strategy as BSMAP except that they align the reads with Bowtie [22] and unique hits are found by a seed-then-extend methodology. (Note that every tool has its own uniqueness criterion. A tool will denote a read to have a unique hit if it finds exactly one occurrence of the read in the reference genome.) Both methods trade accuracy for efficiency.
Interpreting mismatches between reads in base- and color-space. (a,b) Base call error simulation in Illumina and SOLiD reads reflecting one mismatch with respect to the reference from which they are simulated in their respective base- and color-space. (b) A nave conversion of color read to base space, for the purpose of mapping against the base space reference, is not recommended as a single color base error will introduce cascading mismatches in base space. (c) A bisulfite conversion in base space will introduce two adjacent mismatches in its equivalent representation in color space.
SOCS-B [28] and B-SOLANA [29] were developed to map bisulfite color reads. SOCS-B splits a color read into four parts and tries to get hits for any combination of two parts via an iterative Rabin-Karp approach [30]. SOCS-B uses a dynamic programming approach to convert an aligned read to the aligned portion of the reference genome. The conversion starts with all possible four nucleotides as the pseudo-terminal base (rather than just the terminal base from the read). Subsequently, the sub-strings of the four translations are used to generate partial hashing seeds that are then mapped onto the hashed reference genome. However, the running time of SOCS-B is long and the unique mapping rate is too low to be practical. B-SOLANA improves speed and unique mapping rate by aligning against both fully converted and non-CpG converted references simultaneously with Bowtie. The final hits are determined by checking their number of mismatches.
A recent review article [27] reported that Bismark and BS-Seeker are the most recent published methods for mapping bisulfite base reads whereas B-SOLANA is the most recent published method for mapping bisulfite color reads. This review also highlighted the main challenges to develop methods that can map reads unbiasedly and to improve unique mapping rates for mapping color reads.
BatMeth (Basic Alignment Tool for Methylation) was developed by us to address the issues of efficiency and accuracy on mapping bisulfite reads from Illumina and bisulfite color reads from SOLiD. Unlike existing algorithms, BatMeth does not map the bisulfite reads in the initial stage. Instead, BatMeth counts the number of hits of the bisulfite reads to remove spurious orientations of a read. This idea has significantly sped up the mapping process and has also reduced the number of false positives. When dealing with color reads, BatMeth reduced bias on hypomethylation measurements with high initial mismatch scanning. BatMeth also employed a dynamic programming conversion step for the color reads to account for bisulfite mismatch accurately and an incremental processing step to produce higher unique mapping rates and speed (refer to the Materials and methods section for details).
We have compared the performance of BatMeth with recent stable versions of BSMAP (2.4.2), BS-Seeker, Bismark (0.5.4), SOCS-B (2.1.1) and B-SOLANA (1.0) using both simulated and real data sets (BS-Seeker, Bismark and B-SOLANA used Bowtie 0.12.7 in our experiments). With simulated Illumina and SOLiD reads, BatMeth (default mode) recovered the highest number of hits, has the lowest noise rate and is the fastest among the compared programs. BatMeth is also able to produce better unbiased results than the other programs by comparing the detected methylation levels in different genomic contexts over simulated data sets (Illumina and SOLiD reads) of different methylation levels. With a paired-end library, we show the specificity of our Illumina results by counting the pairs of concordant paired reads that fall within the expected insert size of the library. With a directional library, we indicate the specificity of our results with direction-specific information. In summary, BatMeth is an improved bisulfite mapper in terms of speed, recovery rate and accuracy, and, in particular, has addressed the main challenges of mapping color reads identified in [27].
In order to evaluate the performance of our pipeline, we have tested the following programs: BSMAP, BS-Seeker, and Bismark for base-space mapping; and SOCS-B and B-SOLANA for color-space mapping. BS-Seeker and Bismark only output unique hits for each read. BSMAP, SOCS-B and B-SOLANA will output at most one hit per read, with a flag to indicate if a hit is unique. Some reads can map to multiple genomic locations and since a read can only come from one origin, retaining such non-unique mappings will affect the accuracy of downstream analysis such as unbiased methylation site calls. To avoid the problem of wrong methylation calls, all six programs were thus compared with their unique mapping rates.
All our experiments were run on a server equipped with an Intel Xeon E7450 @ 2.40GHz and 128 GB of RAM. We allowed the same mismatch number and CPU threads on all the compared programs in our experiments. Other parameters were kept at default (see Section 1 of Additional file 1 for the choice of parameters used).
We have not included RMAP in our comparisons as it only performs biased mapping in a non-CpG context. MethylCoder was also not included because a newer variant of it, namely B-SOLANA, has been released (MethylCoder's release notes mention that it is now deprecated due to the release of B-SOLANA). BRAT was considered impractical as it only considers one base error in the first 36 bp of a read and therefore was not included in our experiments.
Below, we define 'recovery' to be the portion of the unique hits recovered by the programs. We also define 'accuracy' to be the portion of the recovered hits that are correct. All recorded timings are wall clock times. A 'hit' is a genomic location to which a read is aligned. Lastly, due to sequencing errors and bisulfite mismatches, we allow k (>0) mismatches when mapping a bisulfite read onto a reference. A genomic location is deemed to be unique for a read if it is the only location with the lowest number of mismatches with respect to the read.
We generated 1 million reads, each 75 bp long, which were randomly simulated from the human genome hg19 using the simulator found in RMAP-bs [31]. The data set was built by allowing a maximum of three mismatches per read. Each C in the simulated read, regardless of its context, was bisulfite converted at a uniform rate of 97%. We benchmarked BatMeth and the other methods, BSMAP, BS-Seeker and Bismark, on this data set (see Section 1.1 of Additional file 1 for parameters used). Since the original coordinates in the simulated reads are known, we can evaluate the accuracy of all the programs by comparing their outputs with the original coordinates. We mapped the reads onto the reference allowing at most three mismatches. BatMeth recovered the most number of true positives and the lowest number of false positives and is the fastest program, as shown in Figure 2a.
7fc3f7cf58