MetaProViz::dma, CoRe=TRUE, Log2Distance sign assignment

[dprymidis]

In DifferentialMetaboliteAnalysis.R (lines ~872 and ~875), the sign of Log2(Distance) is currently determined by comparing raw group means (C1 > C2 or C1 < C2). This should be replaced by comparing the absolute group means, as the sign should be reflective of the group with the highest value not the most positive one.

In code:

Now we can adapt the values to take into account the distance

        Mean_Merge <- Mean_Merge %>%
            mutate(`Log2(Distance)` = case_when(
                # If C1>C2 the distance stays positive to reflect
                # that C1 > C2
 872            C1 > C2 ~ paste(`Log2(Distance)` * +1),
                # If C1<C2 the distance gets a negative sign to reflect
                # that C1 < C2
 875            C1 < C2 ~ paste(`Log2(Distance)` * -1),
                TRUE ~ "NA"
                )

This causes issues and it is not reflective of the biology in some cases.
For example:
in C1 vs C2
if C1 = -1 and C2 = -10
Current rule: since -1 > -10, the distance is marked positive
on Volcano: this metaboltie appears on the right side of the Volcano.

Biologically, this is misleading, because the magnitude of C2 (|−10|) is larger than C1 (|−1|).

[ChristinaSchmidt1]

Hi Dimitrios,

Thanks for sharing your thoughts and observations on this case.
Indeed, this is not that easy. For example in the example scenario where we have C1=-10 and C2=1, using the absolute value would mean that C1 > C2 and hence the log2(Distance) is positive. This is wrong as C1 < C2.
This is the reason why we have added additionally the infromation of consumption or release to enable correct biological interpretation, for example:
Case 1:
C1 = -1 and C2 = -10 --> C1>C2 --> +log2(Distance). In addition this will receive the label "Consumed" since both C1 and C2 are < 0.
Case 2:
C1 = -10 and C2 = -1 --> C1 < C2 --> -log2(Disctance). In addition this will receive the label "Consumed" since both C1 and C2 are < 0.

Now if you interpret the volcano plot, you will know how to interpret data points coloured consumed. From the output table you have also access to the mean values of C1 and C2:

If the aim is that the sign (=positive/negative) of the the log2(Distance) should reflect the magnitute (= absolute value difference) for each case, we need to treat each case of a C1 - C2 combinbation differently:

Mean_Merge <- Mean_Merge %>%
            mutate(`Log2(Distance)` = case_when(
            C1 > C2 & C1 <0 & C2 < 0 ~ paste(`Log2(Distance)` * -1), #Both consumed
            C1 < C2 & C1 <0 & C2 < 0 ~ paste(`Log2(Distance)` * 1), #Both consumed
            C1 > C2 & C1 >0 & C2 > 0 ~ paste(`Log2(Distance)` * 1), #Both released
            C1 < C2 & C1 >0 & C2 > 0 ~ paste(`Log2(Distance)` * -1), #Both released
            abs(C1) > abs(C2) & C1 <0 & C2 > 0 ~ paste(`Log2(Distance)` * 1), #C1 consumed and C2 released
            abs(C1) < abs(C2) & C1 <0 & C2 > 0 ~ paste(`Log2(Distance)` * -1), #C1 consumed and C2 released
            abs(C1) > abs(C2)& C1 >0 & C2 < 0 ~ paste(`Log2(Distance)` * 1), #C1 released and C2 consumed
            abs(C1) < abs(C2) & C1 >0 & C2 < 0 ~ paste(`Log2(Distance)` * -1), #C1 released and C2 consumed
            TRUE ~ "NA"
                )

For cases of C1 & C2 <0 or C1 & C2 >0 this would be fine, but in cases of consumption in one condition and release in the other, I do not think that magnitute necessarily reflects biology because a mechanism of release is not necessarily comparable of a mechanism of consumption and hence the magnitute is not either. This is why we opted in first place to use a more simplistic approach of only considering C1 < C2. I will discuss this also with the others, but would be keen to hear your thoughts.