The Computations Tab

1.4	The Computations Tab

The Computations tab has only one function: carry out the computations on the Amplicon Project, with the elements currently defined and the “active” Read Data Set(s). This requires that the Project has been “set up”, including the definition of the various elements that constitute it (Reference Sequences, Amplicons, Read Data Sets / Read Groups, Samples, Variants, and, optionally, MIDs / MID Groups and Multiplexers), and their associations as appropriate. (See sections 1.3 on the Project Tab, and the example in section 2.2, for details on how to set up a Project before computation.)

Figure 1‑44: The Computations tab

Before starting the computation, you have the option of selecting the number of CPUs that will be utilized for the computation. The default value is 1. Increasing the number of CPUs allows certain parts of the computation, mainly trimming and alignment, to be split into concurrent parallel processes, potentially reducing the computation time.

The number of CPUs is selected from the dropdown list above the “Start Computation” button (Figure 1‑44). The dropdown list contains values from 0 through the number of processors available on the Attendant PC or Datarig. The value 0 represents “All available processors” and is equivalent to selecting the highest value in the list. The number of CPUs option is the GUI equivalent to the command line option –cpu, which is described in section 3.3.2.1.

Computations with large references require large amounts of memory. If high numbers of CPUs are used on systems with insufficient memory, the system could freeze due to memory swapping issues.

To carry out the computations, simply click the “Start Computation” button. The Computations Status Table will be populated as the computations progress. The Table comprises 3 columns: Description (of each computational step); Progress; and Status. You cannot enter any information into this Table.

The main steps of a Project’s computations are as follows:

Trim Read Data: for each Read Data Set, the Primer sequences and MIDs, if used, of all the reads are identified for demultiplexing purposes and the trim points are noted for the “Target” sequences. (As mentioned in section 1.1.1.3, trimming the Primers is important because any variations found therein would have no biological significance, and therefore should not be reported by the AVA software.)

Demultiplex Read Data: for each Read Data Set, each read is identified as belonging to one defined Amplicon and assigned to the appropriate Sample, taking into account any relevant MID information. If MIDs are not used, the Amplicon must be associated with one specific Sample, and the read’s Sample is so established. If MIDs are used, the Amplicon is used to determine the relevant Multiplexer associated with the Read Data Set, and then the MIDs found within the read, in conjunction with the MID encoding of Samples defined by the Multiplexer, are used to determine the read’s Sample. Demultiplexing may involve:

splitting the reads of a Read Data Set over multiple Samples, e.g. if the experiment was set up such that one or more Amplicons (which are associated with different Samples either directly, or through the use of MIDs) were present in a PicoTiterPlate Device (GS Junior Instrument) or PicoTiterPlate region (GS FLX+ System) of a sequencing Run; and/or

b.	joining of the reads from multiple Read Data Sets into any given Sample, e.g. if the experiment was set up such that multiple regions of a PicoTiterPlate Device (GS FLX+ System) and/or sequencing Runs contributed reads to the Amplicons that are associated with the Sample.

Align Samples with Reference Sequences: the reads of each Sample are multiply aligned to the Reference Sequences corresponding to the Amplicons with which the Samples are associated. Initially, the reads are aligned with their primers included, but after the reads find their place in the alignment, the primer regions get trimmed off. Using the primers in this way can provide more alignment context and produce more sensitive and accurate alignments at the edges of amplicons (particularly amplicons where there is a sizeable deletion near the target sequence boundaries). Similar Individual reads are grouped into Consensus reads, and the multiple alignments of Individual and Consensus reads are constructed for later viewing in the Global Align and the Consensus Align tabs.

Search for Variants: For each defined Variant, the multiple alignments of the previous step are scanned to determine which Individual and Consensus reads span all the positions of the Variant’s Pattern and, of those that do span all these positions, which reads satisfy all the constraints specified by the Pattern. Statistics are calculated for forward and reverse reads separately, and then pooled together, in order to provide estimates of Variant frequencies. The results of this search are reported in the Variants tab.

In addition, the AVA software automatically searches for potential Variants not explicitly entered into the system. These Auto-Detected “Putative” Variants receive “Intelligent Names” that are unique and compact, yet descriptive (see section 4.2); these Auto-Detected Variants are not automatically loaded into the Project, but can be manually loaded based on particular selection criteria on the main Variants Tab (see section 1.5.2.6). The AVA software automatically searches for substitution (SNP) and block deletion (>=3 bp) Variants. (Insertion Variants and block deletion Variants <3 bp are not currently automatically detected, so manual browsing of the alignments may be needed if these types of Variants are anticipated.) Combined with the ability to selectively load and subsequently sort and filter these Variants based on their frequencies and read-orientation support, the vast majority of interesting Variants can be easily discovered and evaluated from the view of the data provided in the Variant tab’s Sample – Variants Table. By providing the ability to both edit the Variant Status, and filter by that Status, the AVA software provides a simple Discovery Workflow to determine which Variants have been evaluated and what the outcome of that evaluation was. See section 1.5.2.7 for more details on the proposed Variant Discovery Workflow process.

If you modified some information in the Project after computing it (e.g. imported new Read Data Sets, defined new Variants, etc.), you must re-compute it to update the results reported in all the tabs. When you re-compute a Project, the AVA software uses cached results, if possible, for any step that has not changed (except for demultiplexing, which is brief and is always carried out). This can save a lot of time in what would otherwise be needless repetition of calculations.

The Computations tab also has a “Stop computation” button that you can use to abort calculations, e.g. if you decide to make further changes to your Project set up while calculations are on the way. This can be useful for very large Projects, where the calculations can take some time. When you do this, the AVA software accepts the results that have already been re-computed, but it also keeps the results from the previous computation that have not yet been altered by the re-computation. The reason for this is that since Amplicon Projects are incremental, re-calculations are often done after adding Read Data Set(s) or Sample(s) to the Project in a manner such that much of the previously computed results are still valid. On the other hand, the results that were in the process of being re-computed at the time of the interruption could truly be corrupted; the results for these computations, caught in an intermediate state, are removed from the computation’s output (such as in the Variants Tab) and the navigation elements that would be used to load these results are disabled (such as those used to load multiple alignments in the Global Align tab).

Interrupted computations: If a computation (or re-computation) is interrupted, there is a risk that part of the output may not match the state of the saved Project. While the AVA software withholds the potentially corrupted results from the data that was being processed at the time of the interruption, it also maintains the results from previous computations that had not yet been altered at the time of the interruption. Be aware that those older results may not be consistent with more recent updates to the Project. The outcome of this is similar to the case described in the “Caution” at the end of section 1.3 (editing a Project in a manner that is germane to previously computed results). If you find that the data in these tabs does not reflect the current state of the Project, try re-computing it. The only way to be completely sure that the Project is consistent is to allow the computation to run to completion, without interruption.

Before starting the actual computations, the AVA software validates the computationally relevant aspects of the Project setup. This includes most Project elements’ definitions and their associations, such as a Variant pattern and its relationship to the Reference Sequence, or the particular pairings of Samples and Amplicons on particular Read Data Sets. If the software finds any problems, warning messages are displayed, giving the user a chance to address them prior to running the computation. The user may elect to ignore the warnings and proceed; the computation will still run, and shouldn’t throw any errors, but the results may be incomplete because the computation will skip the problematic elements.

Figure 1‑45 shows the different kinds of warning messages that can occur:

•	The warning that the Project has been modified, but not saved, so the computation might produce results that are out of sync with the Project’s current state.

•

The message indicating that further messages in the Computation Warning window are based on the Project in its current state, i.e. as computation would see it if it were saved; this gives you warning of problems in a Project even before you save it to disk. If your Project is up to date on the disk, the other messages below may still occur but they would then concern the Project as saved.

•

A warning that a Read Data Set is active in the Project, but has no associated Amplicons (because no Sample-Amplicon pairs have been assigned to it). This may be an oversight on the part of the user, in which case a large part of the expected output might be missing if the computation were carried out. This warning also appears if a valid Multiplexer is associated with the Read Data Set but no Amplicons are associated with the Multiplexer (and there are no other Amplicons associated with the Read Data Set directly via Samples or indirectly via another Multiplexer).

•	A warning that one or more Amplicons that are associated with some Read Data Set have problems. Such problems may include:

◦	the Amplicon is incompletely defined (no primers, or target start/end not specified)

◦	the Amplicon appears to be inconsistent with its Reference Sequence (the Start/End of the Target are outside the range of the Reference Sequence; this may occur if the user edited the Reference Sequence subsequent to having assigned the Start/End of the associated Amplicon)

◦	the Reference Sequence itself is incompletely defined (e.g. it was given a name, but no actual sequence; in this case the Amplicon wouldn't likely have any Target Start/End set either).

•	A warning that one or more Variants that are potentially associated with a Read Data Set (the Variant location on its Reference Sequences is spanned by an Amplicon that is associated with the Read Data Set) have problems. Possible Variant definition problems include:

◦	the Variant pattern is inconsistent with the Reference Sequence [e.g. a substitute constraint specifies the substituted nucleotide as the same as the one already in the Reference Sequence (should be a match constraint instead)]

◦	some or all the positions of the Variant Pattern are not in the Reference Sequence (as above, this may occur if the user edited the Reference Sequence subsequent to having defined the Pattern of the associated Variant)

•

A warning that the file for an active Read Data Set is missing. This warning would be triggered if a Read Data Set was imported as a symbolic link, and the file that the link points to has been moved, deleted, or become corrupted. The warning message tells where the link was expecting to find the file so that it can be restored to the proper location.

•	A warning that an inconsistent Multiplexer is associated with a Read Data Set. Problems with a Multiplexer may include:

◦	The Multiplexer has not been completely defined (the Encoding, MIDs, and/or Samples have not been specified).

◦	There is a problem with the set of MIDs on either the Primer 1 or Primer 2 sides being used by the Multiplexer to encode Samples. Examples of MID problems are:

▪	An MID is undefined.

▪	At least 2 MIDs are defined and they are not of uniform length (at least one of them has a different sequence length than another).

▪	An MID in the set has an identical sequence to another MID in the set.

Figure 1‑45: A Computation Warning message showing several of the types of message that can occur. The final warning message illustrates that more than one warning for a single Read Data Set can get merged together into a single message. Some of the additional warning messages that can occur, related to Multiplexing, include: Multiplexers that are associated with a ReadData that contain invalid combinations of MIDs, possibly including MIDs that are undefined; or Multiplexers associated with ReadData that have no MID -->Sample associations defined but do have Amplicons associated with the Multiplexer and the ReadData; or the similar case where MID --> Sample associations are defined for the Multiplexer but no Amplicons have been associated with the ReadData-Multiplexer combination.