This is an introduction to estimate genome size and genome character properly. For
more information, you can read the published manuscript.

1. How to finish a simple run?
./gce -f kmer_depth_file >gce.table 2>gce.log ##basic standard discrete model.
./gce -f kmer_depth_file -c unique_depth -H 1 >gce.table 2>gce.log ##for heterozygous mode.
./gce -f Kmer_depth_file -c unqiue_depth -m 1 -D 8 >gce.table 2> gce.log  ##for continuous model

2. How to get a proper genome size?
Firstly, you need to set -g total_kmer_num. You can obtain the value from the output of kmer frequence count programs.
Secondly, you can use the estimate genome size in gce.log.

The genome size estimated here is based the following details:
1) formula: genome_size = now_kmer/final_cvg;
2) now_kmer = total_kmer - low_sum_num( low_sum_num is calculated when updating now_node);
3) final_cvg is much complex, for more detail, please read the published paper.

3. How to decide whether to set -b(for bias data)?

If you count kmer depth with a max value 255 and you need to use a1, then you have to pay attention to this question.
There will be a peak in the depth 255, because when the depth of kmer is larger than 255, they will be added to the 255.
This peak may cause the a1 value to be underestimated especially when there are certain kinds of sequencing bias existing.

When the K size is small, especially less than 19 for human, you do not need to set this parameter. When the K size is larger than 19, you need to set this parameter for real sequencing data.
For E.coli, because the species is quite smaller than 17, you need to set bias for real sequencing data.

4. How to estimate the heterozygous ratio?
When you want to estimate the heterozygous ratio, you need to make sure two points:
1) there is clear half peak in the kmer depth distribution curve(a peak in the unique_depth/2 position).
2) there is no sever bias or sequencing error.

Then you can set -H 1 -c unique_depth. (unqiue_depth is the depth of unqiue peak, you can roughly estimated)
In the gce, we will estimate Kmer Heterozygous Ratio(KHR) based on the formula: KHR = a1/2/(2-a1/2).
To get the Base Heterozygous Ratio(BHR), you can use the formula: BHR = KHR/Kmer(suppose each snp cause K new kmers).

5. How to decide whether use continuous model or discrete model?
The performance of the two model is quite similar. For real sequencing data, continuous model is more suitable, however, its stability should be improved.

6. How to treat the situation when the peak recognized is wrong?
You just set -c , then the peak around the c value will be recognized. This is
quite useful when the main peaks failed to be recognized.

7. How many peaks need to be considered when using continous model?
It is suggested to set -D 8, then the total considered peak number is 48.

Any more questions, please contact: liubinghang@genomics.cn
Thank you!