Avanzado: Argumentos de las acciones de Macro

This section will discuss the features, and issues, of some of the schemes created during the process described in Section 6.2 - a full listing of the schemes is provided in Appendix A. A number of insights emerged from the scheme creation work; however, they are reserved for Chapter 11.

To start the current description, a review of the terminology is provided, and that is used to fuel a discussion on the focus of the schemes. A study of the common characteristics of the schemes finishes this section.

6.3.1

Terminology and focus

The terminology, introduced in Chapter 5.3, shall be used both within the schemes and in the discussion of the schemes:

Experiment The actual research undertaken;

Result The published outcome, i.e. the researcher’s interpretation of the experimen- tal result;

Annotation The (EMAGE or GXD) editors’ analysis of the experiment, including their own interpretation of the result - can be textual or spatial;

Verdict The statement that a gene is (not) expressed in a tissue, which is either stated directly in an annotation or can be derived from an annotation. The term may also be used to refer to the final conclusion drawn by the user.

6.3.2

Portrait of characteristics

The schemes can be classified into three groups: those concerned with the reliability of the underlying experiment; schemes focusing on the reliability of the annotations derived from the experiment; and schemes relating to the anatomy of the mouse.

The first group incorporates a number of subgroups that consider a range of pos- sible issues relating to the notion of a reliable experiment. This list includes: the research team; the method used to conduct the experiment; and, the results obtained

from it. The annotation group likewise has a number of subgroups focusing on spa- tial3 and textual annotations4 in addition to schemes that are applicable to both. The final group, anatomical schemes, considers how the location of the result can be used to infer new knowledge.

The schemes cover a wide range of topics and do so with differing granularity. For example the following scheme can be considered high level:

Experiment E indicates result R Experiment E2 indicates result R Experiment E3 indicates result ¬R

Two experiments indicating the same result, increases the likelihood of the result being correct

Therefore R is probably correct

1. Are E and E2 performed by the same lab? 2. What do other experiments show?

whereas this scheme is more focused:

The spatial annotation SA for experiment E indicates verdict V SA is based on spatial mapping SM

SM features N number of voxels N < 10

Spatial mappings with less than 10 voxels are likely to be errors in the mapping process

SA is not a good indicator that V is true

Ultimately the latter scheme was withdrawn as the expert felt that he could not state with any confidence that it was accurate. To do so would require considerable further research.

3_{The schemes in this group should apply equally to spatial annotations and textual annotations}

derived from spatial annotations.

The second scheme is a rare example of a low level scheme. However, it should be noted that it still does not feature low level biological knowledge such as knowledge of the processes behind gene expression, e.g. transcription. Instead the schemes contain practical guidance in evaluating the information in EMAGE and GXD in such a manner that little biological knowledge is required. The schemes mostly expect that their reader is familiar with the domain vernacular.

A number of schemes have no critical questions. In the case of the second scheme, it is because it was removed before the questions were considered. However, other schemes have no questions because the expert does not think they are necessary, for instance:

ISH experiment E suggests result R

The user, when examining the image from E, has confidence level C in R The image is the actual result of the experiment, and so analysing it gives an

accurate result R is C likely to be true

The expert believes that his peers will be equally as adept at analysing an image as he is, and so he does not see any requirement to question their judgement.

During the creation process a modest number of tangential schemes were produced. For example, one scheme has the conclusion “it is very likely that the experiment used the probe”. This is never integrated into the general topic - how to analyse the contents of EMAGE/GXD. Nor is the scheme retracted as it is an accurate statement. Numerous schemes refer to the experimental result, despite the fact that only the annotation of that result is available. For example:

ISH Experiment E produced result R E has low pattern clarity

Experiments with low pattern clarity are very hard to analyse and so their results are not totally trustworthy

Therefore R is possibly correct 1. What does SAGE show? 2. What does Microarray show?

3. Is the equivalent result true in the previous developmental stage? 4. Is the equivalent result true in the subsequent developmental stage?

In such situations, the intuition behind the scheme can often be applied directly to the annotation. For example, as the annotation is based on the result, any conclusion implying the experiment or its result is not entirely trustworthy must mean that the annotation is not trustworthy. This is captured in:

Annotation A was based on experiment E E cannot be trusted

If an experiment cannot be trusted, an annotation based on it cannot be trusted Therefore A cannot be trusted

A further issue is that a number of schemes cite other resources, e.g. :

Experiment E on the mouse produces result R for gene G and tissue T

Experiment E2 on the zebrafish produces result R for the zebrafish equivalents of G and T

Finding the same result in multiple species provides a very good indication the result is correct

Therefore R is very likely to be correct

This scheme suggests that finding an equivalent result in another species is a good reason to trust a result for the mouse. Implementing this would entail the creation of

a set of schemes to help determine when a mouse gene is equivalent to a gene from another species. A second set of schemes would be required to identify equivalent tissues across the species, with a third set vital to analyse the result of the gene expression experiments in the second species. This is an immense amount of work, and beyond the knowledge of the current expert. Accordingly, this work was not undertaken.

With regard to the argument (scheme) strengths, although the expert had a range of 0 - 100 to choose from, he grouped scores at specific points. An examination of Appendix A reveals that only 10 strengths were used: 100%, 99%, 85%, 80%, 65%, 55%, 50%, 45%, 15%, and 1%. One contributing factor to this effect was that occasionally the expert identified families of schemes that, although subtly different, were based on the same intuition. Such families were assigned the same scores.