DeepConsensus improves the consensus accuracy and yield of the PacBio CCS reads which downstream can improve the quality of genome analysis. To assess the improvement we can achieve by using DeepConsensus reads for genome assembly, we used hifiasm (v0.16.1) assembler to assemble the HG002 sample and used YAK v0.1-r56 to assess the quality of the assembly.

Notably, the DeepConsensus v0.3, v1.0, v1.1 and v1.2 models are trained on the T2T v2.0 assembly of CHM13 (with v1.2 being additionally trained on maize), so HG002 is a held out sample used for this analysis.

We performed genome assembly on reads from two PacBio SMRT Cells with insert sizes of 16kb and 24kb.

Table 1: HG002 assembly metrics with 2 SMRT Cells of PacBio CCS reads with 16kb insert size.

Table 2: HG002 assembly metrics with 2 SMRT Cells of PacBio CCS reads with 24kb insert size.

Here we describe the methods used to perform the assembly and to estimate the quality of the assemblies.

We performed the genome assemblies with hifiasm v0.16.1 assembler using the following command:

hifiasm -o OUTPUT_DIR/OUTPUT_PREFIX -t NUMBER_PROCESSES FASTQ_FILE

We used YAK v0.1-r56 to evaluate the assembly quality. YAK uses the k-mer distribution observed in short-reads (Illumina) to evaluate the quality of the sequence present in the assembly. We used the following command for YAK:

yak qv -t NUMBER_PROCESSES \
-p -K 3.2g -l 100k KMER_DB ASSEMBLY_FASTA > OUTPUT_qv.txt

The data used for this analysis is available publicly.