I wanted to see the effect of coverage on the assembly quality to see at which point there is a diminishing return. I am using paired-end reads only (101x2 with 520 insert size), no mate pairs or long reads. Normally a higher coverage is supposed to increase NGA50, but instead, the contig NGA50 has gone down while the LGA50 has gone up. I have five levels of coverage: 10X, 15X, 20X, 25X and 30X.
15X has the highest contig NGA50 and lowest LGA50, while 30X has the lowest NGA50 and highest LGA50. The order of high NGA50 to low NGA50 is in this order: (Best)15X, 10X, 20X, 25X, 30X(Worst).
The k-mer size used for assembly was 25 bp and was run with SOAPDenovo2 and low-frequency k-mers were not discarded.
I used QUAST to evaluate the assembly.
What explanation(s) could there be for these results?