274

18

Transcriptomics and Proteomics

is practicable up to about 20 cycles and is highly appropriate wherever linear

heteroöligomers sharing a common chemistry are required.

2. For all other cases, minute amounts of the receptor substances are directly

deposited on the array (e.g., using laboratory microrobotics combined with inkjet

technology for applying solutions of the different substances). This is suitable for

large macromolecules, such as proteins, or sets of molecules of substances not

sharing a common chemistry, or longer oligopeptides.

In both cases, each patch is uniquely identified by its Cartesian array coöordinates.

Sample preparation. The raw material is processed to make available the analyte(s)

of interest and possibly partially purified. The mRNA is typically used to generate

a set of complementary DNA molecules (cDNA), which may be tagged (labelled)

with a fluorescent or other kind of label.

Array exposure. The array is flooded with the sample and allowed to reach equi-

librium. Then all unbound sample is washed away. If the analyte was not tagged,

tagging can be carried out now on the chip (e.g., by flooding with a hybridization-

specific dye ⁶) after removing the unbound molecules, which has the advantage of

eliminating the possibility of the tag interfering with the binding.

Array reading. The array is scanned to determine which patches have captured

molecules from the sample. If the sample molecules have been tagged with a fluo-

rophore, then fluorescent patches indicate binding, with the intensity of fluorescence

giving some indication of the amount of material bound, which, in turn, should be

proportional to the amount of mRNA present in the original sample.

Image processing. The main task is to normalize the fluorescent (or other) intensities.

Normalization is important when comparing the transcriptomes from two samples

(e.g., taken from the same tissue subject to two different growth conditions). A

straightforward way of achieving this is to assume that the total amount of expressed

mRNA is the same in both cases (which may not be warranted, of course) and to divide

the intensity of each individual spot by the sum of all intensities. If the transcriptomes

have been labelled with different fluorophores and exposed simultaneously to the

same chip, then normalization corrects for differences in fluorescence quantum yields

and the like.

Analysis. The procedures followed for supervised hypothesis testing will depend on

the details of the hypothesis (Sect. 13.1). Very commonly, unsupervised exploratory

analysis of the results is carried out which, in effect, uses no prior knowledge but

explores the data on the basis of correlations and similarities. One goal is to find

groups of genes that have correlated expression profiles, ⁷ from which it might be

inferred that they participate in the same biological process. Another goal is to group

6 For example, ethidium bromide, the fluorescence of which becomes about 20-fold stronger after

it is intercalated into double-stranded DNA.

7 An expression profile is defined as a two-column table, with conditions in the left-hand column

and the corresponding (relative) amounts of expressed proteins (possibly as RNA) in the right-hand

column.