Table 1.

Quality Assessment Criteria for GPX-MEA

Assessment category	Tool	Weight	(Score)Criteria
Standard Operating Procedures (SOPs)	Experiment description/protocol	2	2 – Standard Operating Procedures have been documented and followed. 1 – Only partial SOPs have been documented applied or followed 0 – SOPs not known or not followed
Sample size	Experiment description/protocol	3	2 – Number of independent biological samples exceeds 5 per group, allowing for reasonably high statistical power in the hypothesis testing. 1- 3–5 independent biological samples per group, allowing to apply minimum level of statistics 0 – 1 or 2 independent biological samples per group (e.g. pilot experiment), reducing the applicability of statistics.
Image quality	Image scan or numerical representation (Affymetrix)	3	2 – No signs of chip defects, background noise (RawQ) within 5 point range of each other. Scaling factor within 1–2 fold range. 1 – Isolated chip defects, RawQ within 5 point range, scaling factor 1–3 0 – Systematic chip defects, saturation, RawQ outside 5 point range, scaling factor greater than 3.
Target sample quality	Electropherogram, UV-Spectrophotometer	1	2 – Presence of 18S and 28S ribosomal subunits with no sign of degradation in any sample (RNA integrity). Absorbance ratio of A260/280 close to 2 (RNA purity) [16]. 1 – Either RNA integrity or purity criteria are met. 0 – Neither integrity nor purity criteria are met, or sample quality assessment was not performed, or no information provided.
Sample level data variation/distribution	Box-and-whisker plots, supporting MA plots if required.	3	2 – Very consistent array medians and Inter-Quartile-Ranges across all samples before normalisation procedures (note: underlying assumption is that any treatments/conditions should not cause differential expression in more than 5–10% of all gene probes on the array.) 1 – Very small number of inconsistent data distributions with assumptions of a correctable difference in array signal intensity met (all genes on array affected, inter-array gene relationships are linear or global differences are expected, e.g. LPS treatment). 0 – Inconsistent median signal levels and spread (IQR) of data throughout the experiment, no biological explanation given.
Gene level data variation	Coefficient of Variation versus Mean Expression plots	2	2 – Coefficient of Variation for majority of genes across all replicates within a sample group is less than 20%. Genes above this CV level are mostly in the lower signal range (due to measurement error). 1 – A small proportion of all genes has higher CV values than 20%, with a small number of these in the medium to high expression range. 0 – A large proportion of genes has a CV above 20% across most of the expression range.

A quality control assessment is made for each experiment curated into GPX-MEA. The quality assessment score is calculated using quantitative and qualitative scores in different areas; use of Standard Operating Procedures, the number and use of independent biological replicates, quality of image, sample quality control and structure of numeric data (spread, reproducibility). A combination of the different areas is used to give a total quality assessment score. This score can be used to indicate the quality of the underlying experiment in the database and as such is subject to interpretation. Where data is unavailable for assessment no score is assigned.