Chromosome-level assembly of Thymus vulgaris

Curro Campuzano Jiménez

2023-06-24

Thyme’s Ecology in Saint-Martin-de-Londres

Figure 1: Spatial distribution of Thymus vulgaris phenolic (black) and non-phenolic (yellow) chemotypes. Obtained from Thompson (2020)

More complete scaffolds, but how?


We have a de novo assembly from HiFi, but it is highly fragmented

  1. Reads as input
  2. Contigs as input

Figure 2: Recently published T. quinquecostatus genome assembly

Main idea


Data exploration

How T. vulgaris reads map to T. quinquecostatus genome?

  • Most windows of T. quinquecostatus (\(\sim 75\%\)) correspond to Zero-process (repetitive sequences?)

\[ Y = \begin{cases} 0, & \textrm{with probability } p \\ \textrm{Poisson}(\lambda), & \textrm{with probability } 1-p \end{cases} \]

Is a homology-based assembly scaffolding approach effective without Hi-C data?

Methods

Reproducible research

  • Chained execution of many CLI
  • Isolated software environment (using Conda and Singularity)


https://github.com/currocam/BiRC-Thyme

Contribution to Snakemake project

Results

Scaffolding process with RagTag (77% placed nucleotides)

Figure 3: Synteny-like plot of CM044164.1 (only four T. vulgaris contigs)

Genome assembly size

Figure 4: Comparison of genome assembly size

Quality assessment: N50 (contiguity)

From 1.87 Mb (n=133) to 48.92 Mb (n=8)

Figure 5: For historical context, N50 statistics of the published plant genomes. Obtained from Sun et al. (2022)

Quality assessment: BUSCO

BUSCO looks for the presence of nearly-universal orthologous genes

Completeness score: 96.4%1 (using Eudicots dataset)

Disclaimer

Complete and duplicated genes is high (31.4%)

  • Recent genome duplication event?

  • Chimeric assembly of haplotypes?

Aligned short reads

Obatined from Herbari Virtual del Mediterrani Occidental

Validation using experimental data

Future work

  • Genome assembly quality evaluation: repetitive sequences

  • Validation with experimental data: transcriptome

  • Improve assembly: alternative reference-based scaffold algorithms

  • Phylogenetic analysis: thyme ecology

Conclussions

Take-home message


De novo assembly of HiFi reads + homology-based scaffolding is an effective approach to obtain a chromosome-level assembly but a more comprehensive evaluation and experimental validation

has to be done

Thanks!

References

Bataillon, Thomas, Perrine Gauthier, Palle Villesen, Sylvain Santoni, John D. Thompson, and Bodil K. Ehlers. 2022. “From Genotype to Phenotype: Genetic Redundancy and the Maintenance of an Adaptive Polymorphism in the Context of High Gene Flow.” Evolution Letters 6 (2): 189–202. https://doi.org/10.1002/evl3.277.
Sun, Yanqing, Lianguang Shang, Qian-Hao Zhu, Longjiang Fan, and Longbiao Guo. 2022. “Twenty Years of Plant Genome Sequencing: Achievements and Challenges.” Trends in Plant Science 27 (4): 391–401. https://doi.org/10.1016/j.tplants.2021.10.006.
Thompson, John D. 2020. “Plant Traits and Ecological Dynamics.” In Plant Evolution in the Mediterranean: Insights for Conservation, edited by John D. Thompson, 0. Oxford University Press. https://doi.org/10.1093/oso/9780198835141.003.0005.

Extra slides

Impact of Long-read sequencing technologies and Hi-C

Figure 6: Evolution of the number of published Chromosome-level genome assemblies. Data was retrieved from NCBI on 6/15/2023.

Hi-C

Obtained from Wikipedia

Quality control of raw reads

  • 23.65 Gb of HiFi long reads

  • 31.34x coverage of T. vulgaris genome

  • Quality control was successful according to FastQC and LongQC

Figure 7: Per sequence GC content. Obtained with FastQC


De novo assembly

  • HiFiasm assembled \(\sim 1.7\cdot 10^6\) long reads into \(\sim1.900\) contigs

  • \(N50= 1.87\)Mb


Important

Highly fragmented, but within the expectation for highly repetitive plant genomes

Bayesian inference using Julia (Turing library)

model.jl
ZeroInflatedPoisson(λ, p) = MixtureModel(
  [Dirac(0), Poisson(λ)], [p, 1 - p]
  )
  
@model function zip(Y)
    α = 1
    θ = mean(y)
    λ ~ Gamma(α, θ)
    pzero ~ Uniform(0, 1)
    Y ~ ZeroInflatedPoisson(λ, pzero)
end;

Whole-genome alignment between T. vulgaris and T. quinquecostatus

Figure 8: Covered areas of T. quinquecostatus in the Whole-genome alignment

Infer gap sizes

RagTag infers the gap size between two adjacent sequences, \(\textrm{seq1}\) and \(\textrm{seq2}\), according to Equation

\[ \textrm{gapsize} = \left(\textrm{aln2}_\textrm{rs} - \textrm{aln2}_\textrm{qs}\right) - \left(\textrm{aln1}_\textrm{re} - \textrm{aln1}_\textrm{qe} + \textrm{len}(\textrm{seq1}\right)) \]

Modelling T. vulgaris long reads aligned to T. quinquecostatus

  1. Align best 5% of reads to T. quinquecostatus assembly and model mapped reads using Bayesian inference
  2. Count mapped reads into 1000-length windows (reduce computational power)
  3. Fit Zero inflated Poisson

\[ Y = \begin{cases} 0, & \textrm{with probability } p \\ \textrm{Poisson}(\lambda), & \textrm{with probability } 1-p \end{cases} \]

Bayesian inference

Figure 9: Sampled posterior distribution of parameters for pseudo-chromosome CM044164.1 using MCMC

Validation using experimental data

Two hybridization conditions to enrich certain sequences: 15% and 5% of nucleotide divergence (Bataillon et al. 2022)

Figure 10: Aligned short Illumina reads to the T. vulgaris assembly

Project in Bioinformatics | MSc. in Bioinformatics at Aarhus University