Thanks much for sending me the histograms by email.

It appears that you data set is very low coverage. If you do not mind me posting one of the histograms here:

I wrote a paragraph below but I now realize that your problem might have a more simple explanation: did you give the reads, or the SOAPdenovo assembly, as input to Kmergenie? Only the former is correct, and I suspect that you did the latter.

Kmergenie expects to see a mixture of distributions in a k-mer histogram constructed from the reads, including one (or two in diploid mode) Gaussian distribution for genomic k-mers. In this histogram, we clearly see a large amount of low-coverage k-mers (with more or less an exponentially decreasing distribution), that Kmergenie assumes to be sequencing errors, but no clear Gaussian distribution that should have appeared at higher abundances. Hence Kmergenie was not able to fit its model to this histogram. The effect is even more significant for the larger k values you tested (25, etc..).

I would try to address:

File "/home/soba/Documentos/Ensambladores/kmergenie-1.6663/scripts/generate_report.py", line 170, in generate_report
    with open(output,'w') as f:
IOError: [Errno 13] Permission denied: 'prueba_report.html'

By:

chown and/or chmod +x /home/soba/Documentos/Ensambladores/kmergenie-1.6663/scripts/generate_report.py

Ok. I failed. Now I am running with one of the two fastaq without assembly which I deleted the bad quality reads<30. After I m going to talk you how the analysis worked.

Hi. I guess that the k-best at 17 it is an artifact. Also there are two peaks one at 42 and other at 72. What do you think about?

Besides I wonder why don't work with the --diploid option.

Other question. I have the genome in two files. Have I to do the analysis with every file? Or I have to collapse both in a single file.

I appreciate much your advice for a newcomer.

./kmergenie /home/soba/Documentos/Ensambladores/mirdeep2/tutorial_dir/SRR653456_2_e.fastq --diploid  -l 15 -k 55  -o prueba2
running histogram estimation
Linear estimation: ~240 M distinct 38-mers are in the reads
K-mer sampling: 1/45
| processing                                                                                         |
[going to estimate histograms for values of k: 55 45 35 25 15
-----------------------------------------------------------------------------------------------------------------------Total time Wallclock  289.346 s
fitting model to histograms to estimate best k
could not fit prueba2-k15.histo
could not fit prueba2-k25.histo
could not fit prueba2-k35.histo
could not fit prueba2-k55.histo
could not fit prueba2-k45.histo
could not predict a best k value
Execution of decide failed (return code 0)

Here the success trial.

./kmergenie /home/soba/Documentos/Ensambladores/mirdeep2/tutorial_dir/SRR653456_2_e.fastq   -l 15   -o prueba2

running histogram estimation
Linear estimation: ~191 M distinct 51-mers are in the reads
K-mer sampling: 1/36
| processing                                                                                         |
[going to estimate histograms for values of k: 81 71 61 51 41 31 21
-----------------------------------------------------------------------------------------------------------------------Total time Wallclock  396.958 s
fitting model to histograms to estimate best k
estimation of the best k so far: 21
refining estimation around [15; 27], with a step of 2
running histogram estimation
Linear estimation: ~296 M distinct 24-mers are in the reads
K-mer sampling: 1/56
| processing                                                                                         |
[going to estimate histograms for values of k: 27 25 23 21 19 17 15
-----------------------------------------------------------------------------------------------------------------------Total time Wallclock  444.631 s
fitting model to histograms to estimate best k
table of predicted num. of genomic k-mers: prueba2.dat
recommended coverage cut-off for best k: 6
best k: 17

You are right. The size is around 25.000kpb. I run the test with -l 15 and -l 29 with a unique merged file. And the second trial is said that the best k is 35.

TRIAL 1

./kmergenie /home/soba/Documentos/Ensambladores/mirdeep2/tutorial_dir/list_filesTY.fastq  -l 15   -o trial_l15
running histogram estimation
Linear estimation: ~379 M distinct 51-mers are in the reads
K-mer sampling: 1/72
| processing                                                                                         |
[going to estimate histograms for values of k: 81 71 61 51 41 31 21
-----------------------------------------------------------------------------------------------------------------------Total time Wallclock  815.1 s
fitting model to histograms to estimate best k
estimation of the best k so far: 21
refining estimation around [15; 27], with a step of 2
running histogram estimation
Linear estimation: ~587 M distinct 24-mers are in the reads
K-mer sampling: 1/112
| processing                                                                                         |
[going to estimate histograms for values of k: 27 25 23 21 19 17 15
-----------------------------------------------------------------------------------------------------------------------Total time Wallclock  895.178 s
fitting model to histograms to estimate best k
table of predicted num. of genomic k-mers: trial_l15.dat
recommended coverage cut-off for best k: 14
best k: 15

-l 15

-l 29 TRIAL 2

./kmergenie /home/soba/Documentos/Ensambladores/mirdeep2/tutorial_dir/list_filesTY.fastq  -l 29   -o prueba5
running histogram estimation
Linear estimation: ~379 M distinct 51-mers are in the reads
K-mer sampling: 1/72
| processing                                                                                         |
[going to estimate histograms for values of k: 81 71 61 51 41 31
-----------------------------------------------------------------------------------------------------------------------Total time Wallclock  504.639 s
fitting model to histograms to estimate best k
estimation of the best k so far: 31
refining estimation around [29; 37], with a step of 2
running histogram estimation
Linear estimation: ~546 M distinct 29-mers are in the reads
K-mer sampling: 1/104
| processing                                                                                         |
[going to estimate histograms for values of k: 37 35 33 31 29
-----------------------------------------------------------------------------------------------------------------------Total time Wallclock  527.335 s
fitting model to histograms to estimate best k
table of predicted num. of genomic k-mers: prueba5.dat
recommended coverage cut-off for best k: 12
best k: 35