Sensitivity is low

[lisanhu]

Previous, because of a program bug, the sensitivity for the human genome is always 0. The reason for the problem is that I forgot about this issue, and we didn't have the upper case conversion patch in the code at that time. This problem is fixed.

Currently, AccSeq could not produce results with more than 95% of sensitivity. Actually, for the yeast reads, the sensitivity number always shows around 51%; for the human genome reads, the number is around 46-47%, which is not acceptable.

Possible reasons:

1. The algorithm parameters are not suitable for a high sensitivity number.
2. The parameters are fine, but the reverse complementary of the reads are not aligned during the process (which occupies about half of the reads).
3. The combination of both factors above.

[lisanhu]

For reason 1. The algorithm parameters are not suitable for a high sensitivity number., we should conduct experiments with many different sets of parameters to see whether we can achieve significantly higher sensitivity with a reasonable set of parameters.

To accomplish this:

1. Redefine the structure of the parameter file, with more reasonable parameters, may result in a change in the algorithm.
2. The current program takes only one parameter file. We need to re-structure the code to be able to consume a set of parameter files, and do the alignment of the same set of reads against all the configurations, and report sensitivity numbers for all these settings. However, this re-structure may not be favored by the users (users will not prepare a set of parameter files and feed into the program when using it), we need to think of a good way to use it as a command line parameter or undo the re-structure in the future.
3. Produce a set of configuration files with a simple script.

[lisanhu]

The voting algorithm of the project is using a hash table, which uses the location as the key. However, there might be shifts in the locations, need to think of a way to tolerate such cases.