In this section, we show the application of the proposed scoring method to identify sexual dimorphism in the IMPC haematological data collected from wild-type (WT) mice, 15–17 weeks in age, over a 3-year period from 1st January 2018 to 31st December 2020, with a minimum required threshold of 50 mice per sex. Our choice of data is inspired by the importance of the haematology parameters in reflecting overall health. The data used in this study can be accessed via the IMPC web portal under the URL www.mousephenotype.org (data release 15.1 – October 2021).
The IMPC is a global effort aiming to generate and characterise knockout mouse lines for every protein-coding gene in mice [21–24]. The IMPC data are collected from several independent centres worldwide [10]. Every centre contributes to the data collection by adhering to a set of standardised phenotype assays defined in the International Mouse Phenotyping Resource of Standardised Screens (IMPReSS - www.mousephenotype.org/impress). Although all centres follow the same Standard Operating Procedures (SOPs), there may be unavoidable or necessary variations in the implementation of the experiments (such as mouse age or time of the day when the test is performed), equipment (such as manufacture, model and kits) as well as the level of expertise and experience of staff, in addition to variations in inbred mouse strain (Table 1) [25]. This may lead to differing results across centres, which makes a universal inference from the results challenging.
IMPC haematology. The IMPC haematology procedure encapsulates 22 measurements of blood properties such as counts and concentrations (white blood cell count, red blood cell count, haemoglobin concentration, platelet counts, etc.), as well as additional and derived haematological parameters (haematocrit, mean red blood cell volume, mean red blood cell haemoglobin, mean red blood cell haemoglobin concentration, etc.). Figure 1 (top) shows red blood cell counts and (bottom) the haemoglobin concentration collected by 11 IMPC centres. The shifts in the means are most likely due to differences in the equipment used to take the measurements and can be removed by normalising the data. The top plot shows consistently higher red blood cell counts in males than females across centres, whereas there is not a clear pattern for the haemoglobin concentration.
Table 1
Mouse strains used by the IMPC centres for the haematological data collected from 1st January 2018 to 31st December 2020.
IMPC centre
|
BCM
|
CCP-IMG
|
HMGU
|
ICS
|
JAX
|
KMPC
|
MRC Harwell
|
RBRC
|
TCP
|
UC Davis
|
WTSI
|
Mouse strain
|
C57BL/6N
|
✓
|
-
|
-
|
✓
|
-
|
-
|
-
|
-
|
-
|
-
|
✓
|
C57BL/6NCrl
|
-
|
✓
|
✓
|
-
|
-
|
-
|
-
|
-
|
✓
|
✓
|
-
|
C57BL/6NJ
|
-
|
-
|
-
|
-
|
✓
|
-
|
-
|
-
|
-
|
-
|
-
|
C57BL/6NJcl
|
-
|
-
|
-
|
-
|
-
|
-
|
-
|
✓
|
-
|
-
|
-
|
C57BL/6NTac
|
-
|
-
|
-
|
-
|
-
|
✓
|
✓
|
-
|
-
|
-
|
-
|
Consensus score. In line with [3], the sexual dimorphism effect is tested for all 22 haematology traits, independently for WT mice from each of the 11 centres, corresponding to the same mouse strain and metadata group split. We used a windowed linear mixed model described in[26, 27] and implemented in the software R [28], packages OpenStats and SmoothWin [29, 30]. As in [3], \(Sex\) and \(Body Weight\) in the fixed effect terms
$$Response=Sex+BodyWeight+e,$$
and Batch (the date when the test is performed on mice) in the random effect term. We then apply the scoring method to obtain a consensus global inference from the multicentre results, following the logic described in the flowchart below (Fig. 2).
Table 2 shows the outcome of the scoring method for the 22 haematological parameters measured by the IMPC, as well as the comparison with a consensus method based on all centres agreeing on a significant sex effect. Using the method proposed here, there is consensus among 11 IMPC centres for 14 traits with \(-\text{log}\left(s\right)>0\), with males on average higher than females for 9 traits (red blood cell count, red blood cell distribution width, haematocrit, platelet count, white blood cell count, lymphocyte cell count, neutrophil cell count, monocyte cell count, eosinophil cell count) and females on average higher than males for 5 traits (mean cell volume, mean corpuscular haemoglobin, mean cell haemoglobin concentration, mean platelet volume, and lymphocyte differential count). For 8 traits, the scoring method leads to zero or negative values, reflecting a lack of consensus (6 traits), or does not reach the minimum threshold of three centres providing measurements for the results to be processed (lack of information in the data − 2 traits).
Table 2
The outcome of applying the scoring method to 22 haematological measurements collected by 11 IMPC centres. The traits are shown in rows followed by the counts for the centre-based statistical test results, the mean effect size for the 11 centres, the consensus score and the inference, which is based on the -log(score) and the sign of the mean effect size. The scoring method identifies consensus in sexual dimorphism across centres for 14 traits (green and red rows), no agreement for 8 traits (blue rows) and 2 traits which do not meet the minimum requirements for the calculation of the score (yellow rows). Only in 2 cases, all centres agree (in bold).
Trait name
|
Count of outcomes across centres
|
Do all centres agree?
|
Consensus score
|
Not significant
|
Male higher
|
Female higher
|
Mean effect size
|
-log(score)
|
Inference
|
Platelet count
|
1
|
10
|
0
|
No
|
1.25
|
0.35
|
Males Higher
|
White blood cell count
|
1
|
9
|
0
|
No
|
1.17
|
1.12
|
Males Higher
|
Lymphocyte cell count
|
1
|
5
|
0
|
No
|
1.01
|
0.86
|
Males Higher
|
Neutrophil cell count
|
0
|
6
|
0
|
Yes
|
0.80
|
1.71
|
Males Higher
|
Monocyte cell count
|
0
|
5
|
1
|
No
|
0.62
|
2.28
|
Males Higher
|
Red blood cell count
|
2
|
9
|
0
|
No
|
0.55
|
0.30
|
Males Higher
|
Red blood cell distribution width
|
1
|
7
|
0
|
No
|
0.53
|
0.74
|
Males Higher
|
Haematocrit
|
4
|
6
|
1
|
No
|
0.38
|
0.16
|
Males Higher
|
Eosinophil cell count
|
0
|
5
|
1
|
No
|
0.35
|
1.08
|
Males Higher
|
Lymphocyte differential count
|
2
|
1
|
3
|
No
|
-0.32
|
0.13
|
Female Higher
|
Mean cell volume
|
1
|
0
|
10
|
No
|
-0.47
|
0.42
|
Female Higher
|
Mean platelet volume
|
1
|
0
|
7
|
No
|
-0.51
|
0.22
|
Female Higher
|
Mean cell haemoglobin concentration
|
3
|
0
|
8
|
No
|
-0.52
|
0.14
|
Female Higher
|
Mean corpuscular haemoglobin
|
1
|
0
|
10
|
No
|
-0.90
|
0.64
|
Female Higher
|
Large Unstained Cell (LUC) count
|
0
|
3
|
0
|
Yes
|
0.83
|
0.00
|
Does not reach the minimum threshold for this analysis
|
Large Unstained Cell (LUC) differential count
|
2
|
1
|
0
|
No
|
0.39
|
0.00
|
Does not reach the minimum threshold for this analysis
|
Neutrophil differential count
|
3
|
2
|
1
|
No
|
0.35
|
-0.07
|
Not enough signal between or across centres to detect SD
|
Basophil cell count
|
1
|
3
|
1
|
No
|
0.25
|
0.00
|
Not enough signal between or across centres to detect SD
|
Haemoglobin
|
5
|
4
|
2
|
No
|
0.13
|
0.00
|
Not enough signal between or across centres to detect SD
|
Monocyte differential count
|
4
|
1
|
1
|
No
|
0.03
|
0.00
|
Not enough signal between or across centres to detect SD
|
Eosinophil differential count
|
4
|
1
|
1
|
No
|
-0.06
|
0.00
|
Not enough signal between or across centres to detect SD
|
Basophil differential count
|
2
|
1
|
2
|
No
|
-0.16
|
0.00
|
Not enough signal between or across centres to detect SD
|