Fleximer Fleximer: Accurate Quantification of RNA-Seq via Variable Length k-mers

Download

git clone https://github.com/chelseaju/Fleximer.git

Installation

Fleximer is implemented in C++11, and requires compiler g++ 4.7 and above. Please run the following commands to compile the executables:

cd Fleximer
make

Binary files are also provided:

sigmer_generation, sigmer_selection, sigmer_count, EM

Data

Fleximer contains two stages: sig-mer preparation and RNA-Seq quantification. In sig-mer preparation, Fleximer takes a predefined partition of transcriptome. The partition divide transcripts into a set of non-overlapping clusters in a specialized fasta format. A regular fasta format contains a header line of transcript ID follow by one sequence:

>seq1
ACGTACGTA

A specialized fasta format contains a header line of a cluster ID and a set of member transcript IDs, follow by member transcript sequence. Each member is separated by '|':

>clusterX|seq1|seq2
ACGTACGTA|ACGAACGTA

The partition can be based on sequence similarity, or functional annotation. By default, we partition the transcriptome based on sequence similarity as described in RNA-Skim. A human transcriptome partition is provided in /Data/reference/human_60mers.fa.

In RNA-Seq quantification stage, Fleximer takes the RNA-Seq data as input. Several examples simulated from polyester are provided in Data/test.

Sig-mer Preparation

Fleximer relies on the structure and properties of suffix tree to discover all the possible sig-mers. It then constructs sig-mers overlap graphs to guide the selection. sigmer_generation takes the specialized fasta file of transcriptome partition as input, and writes all sig-mers to an intermediate file. sigmer_selection takes the output from sigmer_generation and a desired read length to produces a list of representative sig-mers for the quantification stage.

./sigmer_generation Data/reference/human_60mers.fa Data/sigmers.txt
./sigmer_selection Data/reference/human_60mers.fa Data/sigmers.txt 75 Data/selected_sigmer

RNA-Seq Quantification

Fleximer determines the reads origin through the representative sig-mers. It uses the Aho-Corasick algorithm to determine the sig-mers occurrences for each read. For reads that may come from more than one origin, Fleximer uses the EM algorithm to resolve the ambiguity.

./sigmer_count Data/reference/human_60mers.fa Data/selected_sigmer.txt Data/test/len75_sample_46/eClass_count.txt Data/test_len75_sample_46/sample_01.fq
./EM Data/reference/human_60mers.fa Data/test/len75_sample_46/eClass_count.txt 75 Data/test/len75_sample_46/abundance.txt 

The final transcript abundance for each transcript is reported in read count, RPKM, and TPM.