280
LUIZ F. ONUCHIC AND GREGORY G. GERMINO
• Next in importance is the total number of primary recombinants in a
library. In an unamplified library, each recombinant represents a unique
mRNA species. Genes with very low levels of expression (1/50,000 or 1/
100,000) are unlikely to be present in a library containing less than
250,000 recombinants. The number of primary recombinants must be
much higher if one uses expression screening approaches (see below,
Principles ofVector Selection: Expression vs. Non-Expression Libraries).
• Amplification of a library often results in under-representation of some
clones. Unfortunately, there is no accurate way of predicting how many
additional clones must be screened to compensate for this effect. W e
favor initially evaluating the same number of clones as was present in
the unamplified library. If this screen is negative, we recommend making/borrowing a primary library or using a library derived from a different tissue rather than plating an additional number of clones.
• Random-primed libraries often are more representative than oligo(dT)primed libraries. Short oligonucleotides of random sequence can prime
synthesis from multiple sites along an mRNA molecule whereas oligo(dT) usually primes synthesis from only one site. In the latter case,
each mRNA molecule has only one chance of becoming a cloned
cDNA, and rare transcripts are easily lost if any of the steps are inefficient.
It is not uncommon for investigators seeking rarely expressed genes in tissues that express a complex population of genes (i.e., brain, estimated at
>20,000 different expressed genes) to screen more than 1 x 10 6 primary
recombinants. This can rapidly become an expensive and arduous task.
A novel solution to this problern has recently been proposed. This technique
seeks to normalize the number of copies of each gene in the final library
(Patanjali, Parimoa and Weissman 1991). Rather than 3000 copies of genes
A, B and C and only two copies of D, E and F genes as in a conventional
library, a normalized library would ideally have two copies of A- F. In a normalized library, screening of a total of 100 colanies would be required to
ensure a high likelihood of finding a copy of genes A-F whereas over 100fold more clones would be required using a conventionallibrary. Normalized libraries are constructed by annealing single-strand cDNA to a limiting
amount of genomic DNA, removing the excess unbound copies, and cloning only those cDNAs that bind to genomic DNA. In practice, the normalization protocol is only moderately efficient and has the drawback that the
inserts are usually very short.
LUIZ F. ONUCHIC AND GREGORY G. GERMINO
• Next in importance is the total number of primary recombinants in a
library. In an unamplified library, each recombinant represents a unique
mRNA species. Genes with very low levels of expression (1/50,000 or 1/
100,000) are unlikely to be present in a library containing less than
250,000 recombinants. The number of primary recombinants must be
much higher if one uses expression screening approaches (see below,
Principles ofVector Selection: Expression vs. Non-Expression Libraries).
• Amplification of a library often results in under-representation of some
clones. Unfortunately, there is no accurate way of predicting how many
additional clones must be screened to compensate for this effect. W e
favor initially evaluating the same number of clones as was present in
the unamplified library. If this screen is negative, we recommend making/borrowing a primary library or using a library derived from a different tissue rather than plating an additional number of clones.
• Random-primed libraries often are more representative than oligo(dT)primed libraries. Short oligonucleotides of random sequence can prime
synthesis from multiple sites along an mRNA molecule whereas oligo(dT) usually primes synthesis from only one site. In the latter case,
each mRNA molecule has only one chance of becoming a cloned
cDNA, and rare transcripts are easily lost if any of the steps are inefficient.
It is not uncommon for investigators seeking rarely expressed genes in tissues that express a complex population of genes (i.e., brain, estimated at
>20,000 different expressed genes) to screen more than 1 x 10 6 primary
recombinants. This can rapidly become an expensive and arduous task.
A novel solution to this problern has recently been proposed. This technique
seeks to normalize the number of copies of each gene in the final library
(Patanjali, Parimoa and Weissman 1991). Rather than 3000 copies of genes
A, B and C and only two copies of D, E and F genes as in a conventional
library, a normalized library would ideally have two copies of A- F. In a normalized library, screening of a total of 100 colanies would be required to
ensure a high likelihood of finding a copy of genes A-F whereas over 100fold more clones would be required using a conventionallibrary. Normalized libraries are constructed by annealing single-strand cDNA to a limiting
amount of genomic DNA, removing the excess unbound copies, and cloning only those cDNAs that bind to genomic DNA. In practice, the normalization protocol is only moderately efficient and has the drawback that the
inserts are usually very short.