Abnormal k-mer Frequencies in Hybrid Assemblies Using Hifiasm Version 0.19.9

[alex-kozik]

I am working on lettuce genome assemblies. A year ago, I generated a nearly gapless assembly using HiFi and ONT reads with Hifiasm version 0.19.1.

Recently, I started using Hifiasm version 0.19.9 and noticed some issues. Specifically, the k-mer frequency plots in the log files appeared unusual. Additionally, the hybrid assemblies (HiFi and ONT) were highly fragmented with version 0.19.9.

I observed that the reported k-mer frequencies in the hybrid assembly with version 0.19.9 are twice as high as expected. However, when I ran version 0.19.9 on the HiFi dataset alone, the k-mer frequencies matched those produced by the older version (0.19.1).

Based on these observations, I concluded that version 0.19.9 may not be handling HiFi reads correctly in the hybrid assembly approach with ONT. It seems as though the HiFi input file is being read twice.

Below are examples of test runs, and k-mer plots are provided in the attached screenshots.

hybrid (hifi and ont) Hifiasm-0.19.1:
hifiasm_0.19.1 -o Salinas_Asm_D00_Default
--ul KeyGene_ONT_5X_48K_Plus_Sorted.Fasta
Salinas_HiFi_2023_4X_5K_Plus.Fasta
1> Salinas_Asm_D00_Default.1.Log
2> Salinas_Asm_D00_Default.2.Log

hybrid (hifi and ont) Hifiasm-0.19.9:
hifiasm_19.9.r616 -o Salinas_Asm_00X_Default
-ul KeyGene_ONT_5X_48K_Plus_Sorted.Fasta
Salinas_HiFi_2023_4X_5K_Plus_Copy.Fasta
1>Salinas_Asm_00X_Default.1.Log
2>Salinas_Asm_00X_Default.2.Log

simple hifi only Hifiasm-0.19.9:
hifiasm_19.9.r616 -o Salinas_Asm_00W_Default
Salinas_HiFi_2023_4X_5K_Plus_Copy.Fasta
1>Salinas_Asm_00W_Default.1.Log
2>Salinas_Asm_00W_Default.2.Log