choosing l based on heterozygosity

[dcopetti]

Hello, I wonder if there is a way to choose right away the best -l value when working with genomes at different levels of heterozygosity. The genome I am handling (plant, ~340 Mb) is fairly heterozygous (caus_hifasm_l3_stdout.txt ) and I assembled it with three values of -l: and these are the stats of the assemblies:

assembly	total_length	number	shortest	N50	N50n	N70	N70n	N90	N90n
l3.bp.p	361,374,947	67	13,234	37,233,309	5	32,123,771	7	25,947,092	9
l3.bp.hap1.p	345,491,540	113	15,222	32,337,488	5	28,814,549	7	8,655,938	11
l3.bp.hap2.p	341,046,168	91	15,736	30,077,187	5	17,614,102	8	6,036,483	14
l2.bp.p	369,679,714	71	13,234	37,233,309	5	32,123,771	7	13,530,942	10
l2.bp.hap1.p	359,961,961	110	15,222	32,337,488	5	20,789,015	8	10,450,913	12
l2.bp.hap2.p	327,077,977	94	15,736	30,077,187	5	28,814,546	7	8,551,917	12
l1.bp.p	403,965,254	83	13,234	36,517,093	5	28,814,546	8	8,959,895	13
l1.bp.hap1.p	355,422,637	93	15,222	32,337,488	5	20,789,015	8	10,450,913	12
l1.bp.hap2.p	321,495,862	45	18,696	32,764,075	5	19,928,080	7	8,549,279	12

Decreasing from -l3 to -l1 the following trends can be seen:

• the total size of the p_ctg increases
• the imbalance in total size between hap1 and hap2 increases
• Nx values between hap1 and hap2 get closer.

These assemblies are already very good (2n=18, so most chromosomes are in 1-2 contigs already), but I wonder if there is a way to predict which -l value will be more appropriate without running different assemblies. Maybe based on the relative height of the two k-mer peaks? Thanks,

Dario

[chhylp123]

-l1 or -l2 does more conservative purging so that if the heterozygosity rate is higher, some homologous pairs cannot be identified. In this case, hap1 will be larger than hap2. My personal thought is to always use -l3. If the heterozygosity rate is extremely high, have a try with smaller -s. Purging is still a little bit messy...