===== Expected time to see the first results =====

The larger your genome, the longer it will take for //keeSeek// to produce an output. A good starting point for generating primer-like neverwords is the anagram mode with an even distribution of nucleotides (-a 5:5:5:5). Nevertheless, big genomes (e.g. human genome) will require a certain amount of time to produce the first results. Some estimates for different genome sizes are reported in the following table, for both GPU and CPU versions of //keeSeek//. Note that these are just estimates made on our hardware, namely a server without GPU ((server1: Intel Xeon E5540 quad core, 2.53GHz, 32GB RAM)), a server with GPU ((server2: AMD Opteron 6128 quad core, 2.6GHz, 64GB RAM; GPU Nvidia Fermi M2050, 6GB global memory)), a desktop with GPU ((desktop1: Intel Q6600, 2.40 GHz, 3GB RAM; GPU Nvidia GeForce GT 640 GDDR5, 2GB global memory)) and a notebook with GPU ((notebook1: Intel i7 M260, 2.67 GHz, 4GB RAM; GPU Nvidia GeForce 310M, 0.5GB global memory)), and they can widely vary on different machines. The CPU version of //keeSeek// (-N) is discouraged when analysing genomes > 100 MB, and professional GPU cards are recommended.

\\
**Table1. Expected times to produce the first 128 primer-like neverwords using sequential mode (-w 20) on different genomes.**
\\
^ Reference genome              ^ Genome size ^ -w option ^ server1 \\ (no GPU) ^ server2 \\ (with GPU) ^ desktop1 \\ (with GPU)^ notebook1 \\ (with GPU)^
| Mycobacterium tuberculosis    |   4.4 MB    |    20     |    0m01.52s      |     0m1.3s      |    0m0.408s      |   0m23.81s    |
| Amycolatopsis mediterranei    |   10.2 MB   |    20     |    0m33.1s      |     0m1.6s      |    0m1.036s      |    0m54.08s      |
| Arabidopsis thaliana          |   120 MB    |    20     |    6m25.6s     |     0m8.5s     |    0m9.937s      |   7m0.889s      |
| Pyrus sp.          |   500 MB    |    20     |    27m31.2s     |     0m36.3s     |    0m41.847s       |       NOT POSSIBLE    |
| Xenopus tropicalis     |   1.5 GB    |    20     |   1h10m28s      |     1m50.6s     |    2m2.168s       |       NOT POSSIBLE    |
| Homo sapiens                  |   3 GB      |    20     |    2h44m36s     |     3m38.2s     |     NOT POSSIBLE    |       NOT POSSIBLE    |
\\
**Table2. Expected times to produce the first 128 primer-like neverwords using anagram mode (-a) on different genomes.**
\\
^ Reference genome              ^ Genome size ^ -a option      ^ server1 \\ (no GPU) ^ server2 \\ (with GPU) ^ desktop1 \\ (with GPU) ^ notebook1 \\ (with GPU) ^
| Mycobacterium tuberculosis    |   4.4 MB    |    5:5:5:5     |    0m01.4s       |     0m1.1s      |     0m0.612s      |  0m23.28s      |
| Amycolatopsis mediterranei    |   10.2 MB   |    5:5:5:5     |    0m33.1s       |     0m1.8s          |      0m0.868s     |   0m53.99s     |
| Arabidopsis thaliana          |   120 MB    |    5:5:5:5     |    6m23.3s      |     0m8.5s         |     0m9.553s      |   7m2.203s      |
| Pyrus sp.          |   500 MB    |    5:5:5:5     |   27m18.8s      |     0m36.7s     |    0m41.843s       |       NOT POSSIBLE    |
| Xenopus tropicalis     |   1.5 GB    |    5:5:5:5     |    1h10m1s     |     1m50.7s     |    2m1.396s       |       NOT POSSIBLE    |
| Homo sapiens                  |   3 GB      |    5:5:5:5     |    2h44m1s      |     3m09.1s          |       NOT POSSIBLE    |       NOT POSSIBLE    |

\\
**On the importance of reshuffling:** Sometimes you will notice that generation times are unusually long. In that case, changing the seed for reshuffling (-R) could be useful. Try to explore the sequences generated using option -v 2 and look at their variation patterns. If, for example, the last 2 nucleotides are 'A' or 'T', and the reshuffling mask does not affect them, the filters will discard all k-mers ([[http://www.medcomp.medicina.unipd.it/main_site/doku.php?id=keeseek:filters#end_filtering|3' end filtering]]) untill those nucleotides are changed by the lexicographic order of k-mer generation.