Nucleic Acids Research, 2000, Vol. 28, No. 9 E41-e41
© 2000 Oxford University Press
Retro-recombination screening of a mouse embryonic stem cell genomic library
Southern Alberta Cancer Research Center, Departments of Oncology and Biochemistry and Molecular Biology, University of Calgary, Calgary, Alberta T2N 4N1, Canada
Received December 28, 1999; Revised and Accepted March 8, 2000.
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMATERIALS AND METHODSRESULTS AND DISCUSSIONCONCLUSIONREFERENCES |
|---|
Targeted gene disruption is an important tool in molecular medicine, allowing for the generation of animal models of human disease. Conventional methods of targeting vector (TV) construction are difficult and represent a rate limiting step in any targeting experiment. We previously demonstrated that bacteriophage are capable of acting as TVs directly, obviating the requirement for ‘rolling out’ plasmids from primary phage clones and thus eliminating an additional, time consuming step. We have also developed methods which facilitate the construction of TVs using recombination. In this approach, modification cassettes and point mutations are shuttled to specific sites in phage TVs using phage–plasmid recombination. Here, we report a further improvement in TV generation using a recombination screening-based approach deemed ‘retro-recombination screening’ (RRS). We demonstrate that phage vectors containing specific genomic clones can be genetically isolated from a
TK embryonic stem cell genomic library using a cycle of integrative recombination and condensation. By introducing the gam gene of bacteriophage into the probe plasmid it is possible to select for positive clones which have excised the plasmid, thus returning to their native conformation following purification from the library. Rapid clone isolation using the RRS protocol provides another method by which the time required for TV construction may be further reduced. |
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTS AND DISCUSSIONCONCLUSIONREFERENCES |
|---|
Targeted disruption of mammalian genes in embryonic stem (ES) cells and the further generation of mutant animals is a powerful method by which gene function may be characterized. The introduction of positive selectable markers into the gene of interest is facilitated by recombination through flanking regions of homology, subsequently ‘knocking-out’ gene activity. Traditionally, targeting vectors (TVs) have been constructed in plasmids, which have, in many cases, proven to be unstable when attempting to clone certain genomic regions (1). Recent advances in targeting vector construction have not only provided more stability to the TV itself, but have also increased the speed at which TVs may be generated (1). Tsuzuki and Rancourt (1) demonstrated that phage vectors themselves could be used as TVs, eliminating the requirement for ‘rolling out’ plasmids constructed in phage vectors. To further enhance rapid TV construction, it was revealed that positive selection/disruption cassettes could be introduced into phage clones via double crossover recombination with plasmids, circumventing the need to search for unique restriction sites within the phage clone.
Accurate phage–plasmid recombination events have been exploited to improve methods of genomic clone isolation. Initially developed by Brian Seed in 1983 (2), the recombination library screening method takes advantage of in vivo recombination events between phage vectors containing subcloned DNA and a ‘probe’ plasmid containing a region of homology to the gene of interest. In fact, 60 bp of homology appears to be sufficient for phage–plasmid recombination to proceed (2). Seed demonstrated the high selectivity of this method by isolating DNA clones of the ß-globin gene, a member of a large, highly homologous family. In accordance with the results obtained by Shen and Huang (3), Seed (2) reported that a 10% divergence in homology was capable of decreasing the recombination frequency by two orders of magnitude, reducing the levels of aberrant recombination. Despite the efficient recombination and high selectivity of the method, it was still limited by an inability to select for the removal of plasmid sequences from the phage clone.
More recently, single crossover phage–plasmid recombination has been utilized to generate phage TVs which harbor subtle point mutations for more accurate targeted mutagenesis events (4). Single crossovers result in the incorporation of the entire plasmid sequence into the phage, which is an undesired end product for gene targeting purposes. To select for subsequent removal of the plasmid and retention of the desired point mutation, a spi+ (sensitive to P2 interference) negative selection scheme was implemented (4). It was thus determined that this same selection approach may be applied to library screening through phage–plasmid recombination.
The retro-recombination screening (RRS) protocol described in this paper addresses the selection of condensatants by utilizing a negative selectable marker from bacteriophage . Introducing the gam gene into the 
AN13 plasmid (4) allows selection for excision (condensation) of the plasmid sequence via a second recombination event to be carried out on the Escherichia coli P2 lysogen P2392. Presence of the gam gene confers sensitivity to P2 interference (spi+) (5,6), thus restricting the growth of any phage which have not undergone condensation to remove the plasmid sequence. Therefore, phage isolated from the library by RRS are plasmid free and in their native conformation. This allows the smooth progression of conventional restriction mapping required for clone identification, followed by flanking probe isolation essential to the later analysis of targeted alleles. Further, the accuracy of the method is demonstrated by the exclusive isolation of a series of overlapping clones containing the mouse Oct4 gene.
|
MATERIALS AND METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTS AND DISCUSSIONCONCLUSIONREFERENCES |
|---|
Bacterial strains, bacteriophage and plasmid constructs
All bacterial strains used are E.coli K12 derivatives and were chosen due to their relevant genotypes for selection purposes. LE392 (Stratagene, La Jolla) (5) is rec+; supE; supF and was thus used for growing phage under relaxed conditions. The bacterial strain P2392 (Stratagene) is genotypically identical to LE392; however, it is also lysogenic for bacteriophage P2 and is thus immune to subsequent infection by gam+ bacteriophage
(i.e. spi+ phage). LG75 (3) is rec+; supF°; lacZam and therefore requires the presence of exogenous SupF activity for lacZ expression. The E.coli strain MC1061[p3] (2) is rec+; supF° and is a host for recombination plasmids. The p3 episome in MC1061 carries genes for kanamycin, ampicillin and tetracycline resistance; however, both the ampR and tetR genes carry amber mutations and are therefore functional only if SupF activity is provided in trans. MC1061[p3] and LG75 were kindly provided by Dr D.M. Kurnit (University of Michigan, Ann Arbor, MI).
The murine R1 ES cell genomic library was constructed in an altered Syrinx 2A phage vector (7), designated TK (Fig. 2A). The TK replacement phage vector contains amber mutations in three essential genes (Aam, Bam, Sam), preventing the growth of non-recombinant phage (i.e. phage which do not harbor the supF gene from the probe plasmid) in a supF° host. The phage vector is also rap+ and is therefore proficient in phage–plasmid recombination (7). Cloned fragments were generated via partial Sau3a digestion of genomic DNA followed by size fractionation on a sucrose gradient. The stuffer region of the TK vector was removed by XhoI digestion and fragments 12–14 kb in length were ligated by partial fill-in of both phage and vector restriction sites. Recombinants (2.5 x 106) were packaged using a non-irradiated packaging extract (Gigex) and the library was amplified once by isolating phage plated at 40 000 plaques/plate. The TK library was stored at –80°C in 7% DMSO. The representation of genes in the library appears to be comprehensive, as 10 separate known genes have been isolated to date (unpublished results).
|
AN
is a derivative of AN13 (7) and contains the gam gene of bacteriophage (see 4; Fig. 2A). Introduction of the gam gene allows negative selection for the plasmid sequence in P2392. The probe plasmid OCT (Fig. 2B) was created by ligation of a 268 bp region (bp 260–528) from exon 1 of the mouse Oct4 gene (8) into the AN plasmid digested with HindIII and PstI. OCT (Fig. 2C) is essentially the AN13 recombination plasmid (7) containing the Oct4 probe region and was generated by removal of the gam gene from OCT by SalI digestion.
Retro-recombination screening
OCT was electroporated into MC1061[p3] host cells and transformants were selected for on LB-AKT medium (50 µg/ml working concentration each of ampicillin, kanamycin and tetracycline). Positive isolates were used to inoculate 2 ml of LB-AKT-maltose (0.2% maltose) liquid medium and this culture was subsequently used to inoculate a 20 ml LB-AKT-maltose liquid culture, which was grown to an OD600 of 1.0. Cells from this culture were isolated by centrifugation, resuspended in 100 µl of 10 mM MgSO4 and infected with a 100 µl aliquot of the TK library (3 x 108 p.f.u./ml) at a phage:cell ratio of ~1:200. Whole culture lysis and phage–plasmid recombination (Fig. 1A) was allowed to continue in 40 ml NZY-Amp-maltose liquid medium. The presence of only ampicillin in the culture medium is sufficient to select for the maintenance of both the p3 episome and the probe plasmid, yet it also allows the bacteria to grow at a rate such that they may compete with phage turnover. After 6–9 h the remaining cells were lysed with 100 µl of CHCl3 and the phage were isolated in the supernatant, removing cellular debris by centrifugation. The titer was determined on LE392 (~109–1010 p.f.u./ml) and the phage were passaged over LG75 on NZY medium plus Xgal and IPTG (20 mg/ml each). Phage which had integrated the OCT plasmid (and thus the supF gene) via single crossover homologous recombination through the Oct4 probe region (Fig. 1A) were selected for by their ability to form blue plaques on LG75. Isolated positive plaques were placed in SM phage dilution buffer (5) and ~106 p.f.u. were passaged over LE392 in 2 ml NZY-maltose liquid medium. These relaxed conditions allow efficient condensation of the plasmid and recombination through the duplicated region of homology (Fig. 1B). Phage were again isolated in the culture supernatant by centrifugation and phage titer was determined on LE392. Plaques formed on a P2392 bacterial lawn by these phage were isolated in SM phage dilution buffer. Small scale phage DNA preparations and restriction endonuclease digestions were carried out to determine phage clone identity. Southern transfer and analysis with radiolabeled probes were used to provide a more detailed confirmation of the Oct4 clones.
|
|
RESULTS AND DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTS AND DISCUSSIONCONCLUSIONREFERENCES |
|---|
Recombination and retro-recombination
Although the components utilized for recombination screening have evolved over time, the initial steps in the protocol have essentially remained constant (Fig. 1A). For RRS, two key aspects of the protocol set it apart from modifications made in the past. First, the introduction of the
TK library provides a method by which regions of homology for TVs of the phage variety may be rapidly and accurately isolated. Using this method, 3 x 107 phage may be screened simultaneously, resulting in a homogenous population of individual phage clones, already flanked at one end by the TK negative selectable marker. Second, the gam gene of bacteriophage present in the AN recombination plasmid (4) acts as an indispensable negative selection tool for reversion of the phage–plasmid co-integrate (Fig. 1B). This innovation addresses the original problem posed by Seed (2), whose recombination screening protocol lacked a selection scheme for the excision of plasmid sequences, making clone identification a cumbersome task. An alternative selection scheme for phage–plasmid co-integrates which had subsequently lost the plasmid sequence (condensation) was suggested by Perry and Moran in 1987 (9). Their method required plating the recombinant phage on a mixed lawn of bacteria composed of lacZam; supF° and lac–; supF hosts. White plaques arose from condensed phage incapable of suppressing the lacZam mutation, but capable of overcoming their own amber mutations due to the presence of SupF activity in trans. Plaques growing on this lawn were difficult to score, due to their poor growth (10). Another method selected for condensatants on the lacZam; supF° host LG75, which was transfected with a plasmid encoding the phage A and B genes (10). This host-borne plasmid allows the growth of Aam Bam phage which have excised the supF recombination plasmid. The positive white plaques were reported to be easily distinguishable from the blue plaques formed by non-condensed phage; however, in our experience with this method, white plaques are often difficult to isolate due to the large excess of blue plaques on the bacterial lawn. In the RRS protocol, incorporating the gam gene into the recombination plasmid allows revertants to be monitored simply, by plating phage on the restrictive host P2392. Isolated phage may then be characterized by restriction mapping and merely require the introduction of a selectable cassette to be transformed into TVs. We acknowledge that the RRS protocol does not directly assist in the isolation of flanking probes; however, the retrieval of overlapping clones by RRS (Fig. 1C) provides a source of adjacent genomic DNA present in some, but not all, of the ‘incomplete TVs’.
Phage–plasmid recombination
Recombination in vivo between plasmids and phage bearing regions of homology appears to be mainly dependant upon the host recombination machinery, such as the E.coli RecBC pathway and, to a lesser extent, the RecF pathway (3). The gene gam has been shown both biochemically and genetically to be a specific inhibitor of the recBC exonuclease V complex (11,12), reducing the activity of this pathway in its presence. Due to the inhibitory activities of the gam gene on the RecBC pathway, there was some concern that the spi+ selection methods used for detection of precise plasmid excision would have detrimental effects on the initial phage–plasmid recombination step. For this reason, both gam– (OCT) and gam+ (OCT) recombination plasmids were compared for differences in their frequencies of recombination with phage.
As shown in Table 1, the recombination frequency of OCT was approximately one order of magnitude higher than that seen for OCT (2.77 x 10–6 versus 2.53 x 10–7), which is contradictory to expected results. Since Shen and Huang (3) revealed that the RecE pathway is not applicable to phage–plasmid recombination, activation of this pathway cannot account for this difference. This difference may be explained by the observations of Enquist and Skalka (13), where they demonstrated that the late replicative form of bacteriophage is subject to attack by the host recBC exonuclease V. Following the circular ‘theta’ method of replication, switches to a replication mode by which long double-stranded DNA concatamers are formed. Linear duplex DNA such as this is the preferred substrate of exonuclease V. The existence of this replication survival phenotype conferred by the gam gene is supported by the fact that a number of large double-stranded DNA phage encode genes for specific inactivation of the host recBC nuclease (14–16). If this is the case, the gam gene present in OCT may in fact be inhibiting the levels of phage–plasmid recombination through the RecBC pathway. However, it subsequently raises the level of phage survival during late replication above that of the gam– recombinant phage and thus the number of blue plaques seen on LG75 is increased.
|
In their examination of phage–plasmid recombination, Shen and Huang (3) also noted that recombination frequencies were dependent upon the length of homology involved in the process. According to their results, a region of homology 242–315 bp in length should recombine at a frequency of 10–1–10–2. These results, however, were obtained using purified phage clones. The frequency of recombination observed while screening the
TK library was within the range 10–6–10–7, where the region of homology used was 268 bp in length. This decreased level of recombination must therefore reflect the representation of Oct4 clones in the TK library. Compared to the recombination screening frequency of 1.4 x 10–8 for a 700 bp probe reported by Seed (2), this reveals a significant increase in the efficiency of the method.
The condensation event
Following isolation of recombinant phage which had integrated the plasmid sequence, these phage were grown in LE392 to allow for the removal of the plasmid sequence by condensation. Integration of the plasmid into homologous regions of the phage results in a duplication of homology (Fig. 1A), which is then capable of undergoing a second recombination event that restores both the plasmid and phage to their native conformations (Fig. 1B). This event occurs at a frequency of ~1.0% under relaxed conditions (i.e. supF host; D.E.Rancourt, unpublished data), where the SupF activity provided in trans obviates the requirement for the endogenous SupF activity of the plasmid for phage growth. The condensation frequencies were determined for a series of recombinant phage isolated on LG75 (Table 2). The mean frequency of condensation in these experiments was determined to be 2.1%, which is reasonably higher than expected. This is most likely attributable to the relatively large region of homology used for the probe sequence (268 bp), increasing the possibility of a second recombination event within the phage–plasmid co-integrate.
|
Preliminary data (not shown) also suggest that the condensation event need not be carried out under relaxed conditions in the host LE392. If recombinant phage from the blue plaques on LG75 are immediately passaged over P2392, a small number of plaques form, which represent gam– phage. This suggests that during replication of the phage in LG75, condensation may occur amongst individual or between multiple phage within the linear concatamers, thus resulting in the restoration of the phage library inserts through precise excision of the plasmid. The SupF activity provided in trans by non-condensed phage permits growth of the supF° phage, as well as lacZ expression in the host. Non-condensed phage from this plaque are then subject to negative selection on the host P2392. It appears that the relaxed conditions simply allow a larger number of phage to undergo condensation. Therefore, in situations where larger regions of homology are used as probes the RRS protocol could be reduced in length accordingly, whereas with shorter probes, passage of the phage–plasmid co-integrates over LE392 may be required.
Analysis of isolated phage clones
Oct4 was chosen as a model gene to demonstrate the efficiency of the RRS method in accurate clone isolation. Following passage of the phage over P2392 to select for single crossover condensatants, a subset of the isolated plaques obtained were picked and their DNA prepared. Restriction analysis of the phage DNA was used to confirm the identity of the genomic clones, which are in agreement with the map generated by Yeom et al. (8; Fig. 3A). Initial digestions consisted of removing the TK arms via SalI digestion and simultaneous digestion with HindIII to fragment the cloned region. Figure 3A depicts the results of such a digest. In all cases, the phage DNA was also digested with EcoRI and BamHI, either alone or in concert with SalI, to ensure that the cloned genomic region was legitimate (data not shown).
|
To demonstrate that the expected orientation of the cloned genomic region was correct, Southern blots of the HindIII/SalI-digested phage DNA were probed with radioactively labeled fragments representing the 1.5 kb herpes simplex virus (HSV) thymidine kinase (TK) gene and the 20 bp 3' and 5' Oct4 primers used initially to isolate the recombination probe via RT–PCR (Fig. 3B and C). The TK probe assisted in the determination of orientation, since proximal genomic fragments were found to be linked to the TK gene under the digestion conditions used. The 3' and 5' probes were of use since they bridged a HindIII site present within Oct4 exon 1 (8). This allows the identification of two separate bands, as well as some confirmation that the cloned genomic region represented the Oct4 gene. Importantly, the 268 bp fragment used in the RRS protocol also hybridized to the cloned region of all five phage (data not shown), indicating that the genomic clones contain, at least in part, the 5'-UTR region of the Oct4 gene. Additional hybridizations, including an attempt to identify plasmid sequences by probing with labeled, linearized
AN revealed no detectable signals (data not shown).
The maps of five clones obtained using the RRS method are summarized in Figure 3C. A number of these clones (nos 1–3) were nearly identical and only one phage (no. 5) was found to contain a genomic insert in the opposite orientation. All of the clones tested covered the entire Oct4 coding region, with an average insert size of ~14.1 kb. Since the genomic library was generated in the TK vector, essentially all clones are TVs of the phage variety following the introduction of a positive selectable marker (1). Thus, the RRS protocol allows the isolation of TVs for specific gene knockouts. The largest clone obtained was estimated at ~15.3 kb, therefore allowing a positive selectable disruption cassette of up to 3.4 kb in length to be inserted, without pushing the phage length over the upper packaging limit [105% of the wild-type genome length (48.5 kb) = 50.9 kb]. Most importantly, the fact that the only phage isolated were those containing the Oct4 gene lends support to the accuracy of the RRS protocol in the isolation of these TVs.
|
CONCLUSION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTS AND DISCUSSIONCONCLUSIONREFERENCES |
|---|
RRS provides a means by which phage clones for specific genes from a recombination-proficient library may be accurately and rapidly isolated. However, the method is not applicable to gam+ phage vector libraries such as
gt11. Here, 3 x 107 phage bearing ~12–14 kb inserts have been screened simultaneously, resulting in the amplification and selection of a series of clones bearing inserts which overlap the same genomic region. Probe fragments of a variety of lengths may be used, such that only a small sequence of the gene of interest need be known prior to screening. It has been demonstrated that 60 bp of homology is sufficient to allow recombination to occur (2) and recent experimentation with the RRS method has provided recombinant phage using a 50 bp region of homology at a frequency ~10–2–10–3-fold lower than that observed with the 268 bp Oct4 probe (K.Woltjen, unpublished data). The massive reduction in recombination efficiency with slight decreases in sequence homology (2,3) provides a selection by which only bona fide clones recombine with the probe sequence, amplifying them in the population. By using the gam gene as a negative selectable marker, it is possible to isolate phage clones which have reverted to their native conformation, thus allowing restriction analysis of the clones to proceed smoothly. The data provided above also suggest that the gam gene may convey a survival capability to the phage which is not available in other recombination screening systems. The entire retro-recombination procedure can be completed in 4 days under optimum conditions. As well, preliminary data suggest that the passage of recombinant phage through LE392 may not be necessary, reducing the protocol time to 3 days. The possibility of bypassing the relaxed conditions step, however, may be a function of probe fragment length, such that short probe sequences may not spontaneously give rise to significant numbers of condensed phage clones. This method, therefore, further decreases the amount of time required to isolate a region of homology necessary for the generation of a knock-out TV. Combined with the double crossover cassette insertion method (1), RRS allows TVs for gene disruption to be prepared within 1–2 weeks.
|
ACKNOWLEDGEMENTS |
|---|
We would like to acknowledge Gilbert Schultz for providing the Oct4 fragment which was used as a homologous probe to screen the library, as well as the 3' and 5' primers which were used as probes for Southern analysis. The R1 ES cell line from which the
TK genomic library was generated was kindly provided by Andras Nagy. We would like to thank Todd Unger for his assistance in the creation of the figures and for providing the AN recombination plasmid. Also, we thank Teruhisa Tsuzuki, Frans van der Hoorn and Karl Riabowol for critical reading of this manuscript. This work was funded by the Alberta Cancer Board and Medical Research Council of Canada. K.W. is funded by studentships from the National Science and Engineering Research Council of Canada and the Alberta Heritage Foundation for Medical Research. D.E.R. is a scholar of the Alberta Heritage Foundation for Medical Research.
|
FOOTNOTES |
|---|
* To whom correspondence should be addressed. Tel: +1 403 220 2888; Fax: +1 403 283 8727; Email: rancourt@ucalgary.ca Present address: Gerard Bain, Aventis Pharmaceuticals Inc., 26 Landsdowne Street, Cambridge, MA 02139-4234, USA
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTS AND DISCUSSIONCONCLUSIONREFERENCES |
|---|
-
1 Tsuzuki,T. and Rancourt,D.E. (1998) Nucleic Acids Res., 26, 988–993.
2 Seed,B. (1983) Nucleic Acids Res., 11, 2427–2445.
3 Shen,P. and Huang,H.V. (1986) Genetics, 112, 441–457.
4 Unger,M.W.T., Liu,S.Y. and Rancourt,D.E. (1999) Nucleic Acids Res., 27, 1480–1484.
5 Sambrook,J., Fritsch,E.F. and Maniatis,T. (1989) Molecular Cloning: A Laboratory Manual, 2nd Edn. Cold Spring Harbor Laboratory Press, Cold Spring Harbor, NY.
6 Haggard-Ljungquist,E., Barreiro,V., Calendar,R., Kurnit,D.M. and Cheng,H. (1989) Gene, 85, 25–33.[Web of Science][Medline]
7 Lutz,C.T., Hollifield,W.C., Seed,B., Davie,J.M. and Huang,H.V. (1987) Proc. Natl Acad. Sci. USA, 84, 4379–4383.
8 Yeom,Y.I., Ha,H.-S., Balling,R., Scholer,H.R. and Artzt,K. (1991) Mech. Dev., 35, 171–179.[Web of Science][Medline]
9 Perry,M.D. and Moran,L.A. (1987) Gene, 51, 227–236.[Web of Science][Medline]
10 Kurnit,D.M. and Seed,B. (1990) Proc. Natl Acad. Sci. USA, 87, 3166–3169.
11 Unger,R. and Clark,A.J. (1972) J. Mol. Biol., 70, 539–548.[Web of Science][Medline]
12 Unger,R., Echols,H. and Clark,A.J. (1972) J. Mol. Biol., 70, 531–537.[Web of Science][Medline]
13 Enquist,L.W. and Skalka,A. (1973) J. Mol. Biol., 75, 185–212.[Web of Science][Medline]
14 Tanner,D. and Oishi,M. (1971) Biochim. Biophys. Acta, 228, 767–769.[Medline]
15 Wackernagel,W. and Herrmanns,U. (1974) Biochem. Biophys. Res. Commun., 60, 521–527.[Web of Science][Medline]
16 Sakaki,Y. (1974) J. Virol., 14, 1611–1612.
CiteULike 
Connotea 
Del.icio.us What's this?
This article has been cited by other articles:
|
|
 |
|
  F. Mansergh, N. C. Orton, J. P. Vessey, M. R. Lalonde, W. K. Stell, F. Tremblay, S. Barnes, D. E. Rancourt, and N. T. Bech-Hansen Mutation of the calcium channel gene Cacna1f disrupts calcium signaling, synaptic transmission and cellular organization in mouse retina Hum. Mol. Genet., October 15, 2005; 14(20): 3035 - 3046. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 P. Delmar, S.ép. Robin, and J. J. Daudin VarMixt: efficient variance modelling for the differential analysis of replicated gene expression data Bioinformatics, February 15, 2005; 21(4): 502 - 508. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||