Figures
Abstract
Rare (RVs) and common variants of the RET gene contribute to Hirschsprung disease (HSCR; congenital aganglionosis). While RET common variants are strongly associated with the commonest manifestation of the disease (males; short-segment aganglionosis; sporadic), rare coding sequence (CDS) variants are more frequently found in the lesser common and more severe forms of the disease (females; long/total colonic aganglionosis; familial).
Here we present the screening for RVs in the RET CDS and intron/exon boundaries of 601 Chinese HSCR patients, the largest number of patients ever reported. We identified 61 different heterozygous RVs (50 novel) distributed among 100 patients (16.64%). Those include 14 silent, 29 missense, 5 nonsense, 4 frame-shifts, and one in-frame amino-acid deletion in the CDS, two splice-site deletions, 4 nucleotide substitutions and a 22-bp deletion in the intron/exon boundaries and 1 single-nucleotide substitution in the 5′ untranslated region. Exonic variants were mainly clustered in RET the extracellular domain. RET RVs were more frequent among patients with the most severe phenotype (24% vs. 15% in short-HSCR). Phasing RVs with the RET HSCR-associated haplotype suggests that RVs do not underlie the undisputable association of RET common variants with HSCR. None of the variants were found in 250 Chinese controls.
Citation: So M-T, Leon TY-Y, Cheng G, Tang CS-M, Miao X-P, Cornes BK, et al. (2011) RET Mutational Spectrum in Hirschsprung Disease: Evaluation of 601 Chinese Patients. PLoS ONE 6(12): e28986. https://doi.org/10.1371/journal.pone.0028986
Editor: Amr H. Sawalha, University of Oklahoma and Oklahoma Medical Research Foundation, United States of America
Received: August 26, 2011; Accepted: November 17, 2011; Published: December 9, 2011
Copyright: © 2011 So et al. This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Funding: This work was supported by research grants from the Hong Kong Research Grants Council HKU 765407M to M-MG-B and HKU 778610M to PK-HT; and from The University of Hong Kong Seed Funding Programme 200611159028 to M-MG-B. Support was also obtained from The University of Hong Kong Genomics Strategic Research Theme. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Competing interests: The authors have declared that no competing interests exist.
Introduction
Hirschsprung's disease (HSCR) is a developmental disorder characterized by the absence of ganglion cells in lower digestive tract. Aganglionosis is attributed to a disorder of the enteric nervous system (ENS) whereby ganglion cells fail to innervate the lower gastrointestinal tract during embryonic development. The disease can be classified according to the length of the aganglionosis into short segment (S-HSCR; aganglionosis up to the upper sigmoid colon; 80% of HSCR cases), long segment (L-HSCR; aganglionosis up to/beyond the splenic flexure), and total colonic aganglionosis (TCA) forms. HSCR can manifest isolated or syndromic. There is significant racial variation in the incidence of the disease, and it is most often found among Asians (2.8/10,000 live births). HSCR usually presents sporadically (80% of the cases), and is most frequent in males (M∶F = 4∶1). Familial HSCR shows a non-Mendelian inheritance pattern [1].
Both rare and common germ-line RET variants play a role in HSCR. While RET common variants are strongly associated with the commonest manifestation of the disease (male, S-HSCR and sporadic forms), rare coding sequence variants (CDS-RV) are more frequently found in the less common and more severe forms of the disease (females, L/TCA-HSCR and familial) [2]. These CDS-RVs have reduced, sex-dependent penetrance and account for ∼50% of the familial and for ∼20% of the sporadic cases. Over 200 RET CDS-RVs (www.hgmd.cf.ac.uk) have been described in HSCR. These HSCR RET CDS-RVs mainly cause loss-of function of the RET protein, a tyrosine kinase receptor. Critically, RET gain-of function CDS-RVs are directly implicated in hereditary thyroid cancers [3].
Here we present the screening of the RET CDS-RVs in 601 Chinese HSCR patients, the largest number of patients ever reported.
Results
In this study we report those RET CDS-RVs (including intron/exon boundaries) that were not found in our controls. The overall frequency is <1%. Exception is R114H, which is present in 38 patients (6.32%) but no in controls. R114H is a founder mutation within Chinese HSCR patients [4].
RET variants identified in 601 Chinese HSCR patients
We identified 61 different heterozygous RVs distributed in 100 HSCR patients (16.64%) (Tables 1 and S4, Tables S1(A)-S1(B)). These include 14 silent, 29 missense, 5 nonsense, 4 frame-shifts, and one in-frame amino-acid deletion in the CDS, 3 deletions and 4 nucleotide substitutions in the intron/exon boundaries and 1 nucleotide substitution in the 5′ untranslated-region (UT).
Five variants (R114H/R313Q/V292M/P841P/R180X) are in the dbSNP and one (T278N) in both dbSNP and 1000 Genomes Project. No population/frequency was reported (March/2011). Excluding R114H, no parental DNA was available to determine whether those “database-reported” variants identified in our study had been inherited from unaffected parents or arisen “de novo”.
We identified 6 patients with de novo events, 5 with maternally inherited and 6 with paternally inherited variants. Four patients harboured 2 CDS-RVs.
Excluding those patients included in the GWAS, whole RET deletion in patients homozygous for RET SNPs was not evaluated.
Rare variants in other HSCR series
While 50 RVs identified in this study are novel, 6 (R180X/T278N/R313Q/E480K/Y1062C/M1064T) have been reported in other HSCR patients (Chinese/other ethnicities) and 2 (R114H/V292M) in individuals with other RET related disorders (Material S1).
Y1062C and M1064T have been extensively studied [5]. While Y1062C affects the RET downstream signalling no RET disruption has been demonstrated for M1064T [6]. This is in line our bioinformatics prediction (Table S2) whereby Y1062C is predicted damaging and M1064T benign. The patient reported here harbouring Y1062C (C16C_Male_L-HSCR) was born to unaffected parents and had a distant relative affected. C16C also has V397M in cis. We could not determine whether these variants are de novo, yet the fact that Y1062C has been reported advocates for an inherited mutation that segregates with reduced penetrance. The recurrence of RET CDS-RVs in ethnically different patients suggest that these events occurred before the European-Asian split.
Knowing whether RET de novo variants recur or, whether an amino-acid residue is more prone to DNA changes should be useful in the identification of putative RET mutation hot-spots. R180 residue (CGA) has repeatedly been altered leading to different types of mutations (R180X [7]; R180P [8] and R180Q [9]). The same is true for R114 and T278. Specific RET residues are inherently more prone to mutations than others.
RET coding sequence rare variants (CDS-RVs) distribution across the different HSCR subtypes
The distribution of the RET CDS-RVs was analysed across different HSCR categories. HSCR patients were therefore stratified according to i) length of the aganglionic segment -severity of the HSCR phenotype-; ii) presence/absence of associated anomalies/syndrome; iii) presence/absence of familial involvement.
Length of the aganglionic segment.
When patients were stratified according to the length of the aganglionic segment, the highest frequency of CDS-RVs was found among patients with the most severe forms of the disease (Table S1A). This is in accordance with our previous report [2]. Likewise, the highest frequency of CDS-RVs was found among females, and this became more obvious as the severity of the HSCR phenotype increased (Table 1 and Tables S1A and S1B).
We then investigated whether there was any correlation between type of RET mutation and severity of the HSCR phenotype: among the 17 “mutated” patients affected with the most severe phenotype, 8 (47.07%) had missense changes or in-frame deletions; 5 (29.41%) had non-sense or frame-shifts; 2 (11.76%) had silent variants and 2 (11.76%) had intronic substitutions. Among the 59 “mutated” S-HSCR patients, 48 (81.36%) had missense changes; 3 (5.08%) had non-sense or frame-shifts; 6 (10.17%) had silent changes, and 2 (3.39%) had variants in the intronic or 5′UT regions (Table S3).
Distribution of the coding sequence rare variants (CDS-RVs) across the RET gene
It is assumed that HSCR RET CDS-RVs are scattered all over the gene. Yet, some reports indicate that HSCR RET CDS-RVs tend to cluster in the extracellular domain of the protein [8], [11], [12]. “Exonic” variants (N = 55) where in excess in the sequence encoding the extracellular domain (EC) of the protein. Thirty-six (65.45%) variants were in the EC; 3 (5.45%) in the trans-membrane domain and 16 (29.09%) in the intracellular domain (Table S3).
Exon 5 (cadherin-domain) was the exon with more variants (N = 9) followed by exon 3 (N = 7). Similar data have been reported by others [8] and whether this is related to the high recombination rates reported for this exon needs further investigation [2].
The distribution of variants across RET is relevant as the functional consequences of missense CDS-RVs correlate with their position in the sequence [5], [13], [14]. However, after over 200 RET deleterious variants reported, no correlation could ever be established which reflects the contribution of other genetic factors to the disease including that of HSCR-associated RET common variants/SNPs.
Assessment of the functional effect of the RET variants
Missense CDS-RVs were assessed for their effect in the protein. Benign missense CDS-RVs, silent and intronic variants, were scrutinized for interference with i) signature elements that govern the splicing machinery (exonic/intronic splicing enhancer/silencer sequences -ESE/ESS-), and ii) RNA stability, structure and folding (Table S2).
Out of 29 missense variants identified, 10 were predicted damaging by both Polyphen and SIFT, 16 were predicted damaging by Polyphen and tolerated by SIFT and 3 tolerated by both software packages.
Except L465L, all silent variants were predicted to hinder splicing. Similarly, all benign non-synonymous/missense variants as well as the “in-frame” deletion G731del were predicted to interfere with the splicing machinery. Four intronic variants were predicted to affect splicing by either disrupting the splice sites (acceptor/AS or Donor/DS) or the branch point (BP). The 5′UT variant did not alter any regulatory site.
As the effect of point mutations may be dependent on conservation at both DNA and protein levels, we produced conservation scores for each variation site (Table S2). Seventeen DNA sites had the maximum score implying that changes on those sites are not tolerated. At protein level, the most evolutionary conserved amino-acids are 4 residues with silent changes (de novoG954G/L651L/P841P/L1077L).
Silent variants with the highest conservation scores at DNA (Q327Q/L465L/L1077L) and/or amino-acid level (L651L/P841P/G954G) and non-synonymous/missense that overlapped with functional RNA structures (F961L/Y1062C) according to the UCSC Evofold track were evaluated for their impact on RNA secondary structure/stability. All but Q327Q were found to alter the wild-type RNA stem loop (Figure S1). The only silent variant with no predicted “damaging” effect thus far (L465L) was found to alter such RNA structure.
Bioinformatics prediction indicates that all variants identified but one (c.-37G>C) potentially affect the RET function in a quantitative and/or qualitative manner.
RET rare variants are not necessarily on the RET HSCR-risk haplotype
A common RET-haplotype highly associated with HSCR comprising a functional variant (rs2435357 C/T; T = HSCR associated allele) in intron 1 is associated with reduced expression of RET both in vitro and in vivo [2], [15], [16], [17], [18]. The CDS-RVs cis/trans location with respect to this low-expressing risk-haplotype is relevant to the effect and/or penetrance of the variant as the amount of mutated protein synthesized (hence the effect) would depend on the haplotype where the CDS-RVs resides. The CDS-RVs cis/trans location could also explain the RET SNPs association with HSCR by means of synthetic association [19]. Out of the 100 HSCR patients with CDS-RVs unique to HSCR (Table S4), 68 were homozygous for the RET-risk-haplotype (TT), 9 were homozygous for the wild-type counterpart (CC) and 23 were heterozygous (CT). Of these 23 heterozygous patients, we were able to phase their mutation with the haplotype for 15 (Material S1) [2], [17], [20]. There is not excess of risk-haplotype among patients with CDS-RVs when compared to the whole patient group (χ2 = 0; p = 0.99), indicating that the occurrence of RET CDS-RVs is independent of the haplotype background. RET CDS-RVs do not seem to contribute to the strong association of the RET SNPs with HSCR.
Discussion
This is the largest RET sequencing study of HSCR patients ever conducted. Its aim is to provide a catalog of the RET coding sequence rare variants present in Chinese HSCR patients and assess the distribution of those variants across the different types of HSCR patients. In line with previous observations [2], we found that RET CDS-RVs are more frequently found in the less common and more severe forms of the disease (females, L/TCA-HSCR and familial). As in many other diseases, one would expect a correlation between phenotype and location of the causal variant in the gene and this was not the case. Similarly, the phenotypic variability observed among patients bearing the same variant stresses the difficulty in establishing genotype-phenotype correlations. Undoubtedly, other genetic factors affect the penetrance of RET rare variants and contribute to the phenotypic variability [10], [20].
We used bioinformatics tools to predict the possible implications of these variants on the gene function. Importantly, all variants (but one) were predicted to affect the protein function at different levels. The prediction of missense mutations should perhaps be the most straight forward, yet, discrepancies between Polyphen and SIFT seem to be the norm rather than the exception [21]. More daunting is to assess the predictions provided for silent and/or regulatory variants. However, bioinformatics claims cannot categorically be dismissed no fully accepted and this is widely exemplified in the literature. For instance, 19 genes were found to have truncating variants in controls in the resequencing of X-chromosome exons in mental retardation [22]. On the other hand, silent and/or regulatory RVs are often neglected and these variations underlie as many as 50 diseases [23]. Functional validation and/or faithful segregation of variants with the disease phenotype (which would only be achievable among the scarce familial cases with fully penetrant variants) are needed to fully assess the impact of those variants on the phenotype. Thus, in spite of the ambiguity regarding the potential pathogenicity of many of the variants found and based on the indisputable role of RET in HSCR we suspect that all RVs found are damaging. With a few exceptions, the rare RET RVs presented here are unique to HSCR patients and the existence of some predicted deleterious variants in the general population (1000 Genome project) only confirms that additional genetic/environmental factors are in some instances needed for the disease to occur. As the majority of causal RET variants described thus far are heterozygous it would be tempting to speculate that the manifestation of the disease may be dependent on factors affecting the expression of the RET allele in which the variant resides. We have phased the rare variants with the RET risk-haplotype and conclude that the undisputable association of RET common variants with HSCR is not due to rare variants that went undetected in our previous GWAS study. Thus, this indicates that common and rare variants may contribute to the disease independently although it does not disregard a possible joint effect as stated above. Exhaustive functional analysis of these variants would be out of the scope of this study and although it could back up the bioinformatics prediction “in vitro” data alone will not ensure casualty. This study conveys the problematic assessment of the role of rare variants in disease and emphasises the difficulty in establishing genotype-phenotype correlations. Yet, providing a mutation profile for such a large number of patients will eventually help elucidate the architecture of the disease.
Materials and Methods
The study was approved by the institutional review board of The University of Hong Kong together with the Hospital Authority (IRB:UW06-349/T/1374). Blood samples were drawn from all participants after obtaining informed consent (parental consent in newborns and children below age 7).
Patients and controls
A total of 601 HSCR patients (Males = 490; Females = 111; M∶F = 4.4∶1) recruited throughout Hong Kong and Mainland China were included in the study. Four patients (C16C/C48C/HKC2C/HK75C) reported having a distant relative with the disease, but never parents. Diagnosis was based on histological examination of either biopsy or surgical resection material for absence of enteric plexuses. Sixty-seven patients were syndromic and among those, 16 had Down syndrome (Table 1).
Among the 601 HSCR patients, 430 patients had been genotyped for 21 RET SNPs across a ∼60 kb region of the RET gene as we previously described [4], 258 had been included in our previous genome-wide association study (GWAS) on Chinese HSCR (discovery phase/replication) [20] and 192 in the International Hirschsprung disease study on the differential contributions of RET mutations to Hirschsprung disease liability [2]. The RET rare variants for 86 of the 601 HSCR patients had been previously reported [12]. As controls, 250 ethnically matched individuals (Males = 160; Females = 90) with no diagnosis of HSCR were included.
Sequencing of the RET coding regions
The RET 21 exons were screened in patients, parents when available, and controls to avoid the ascertainment bias caused when only the exons with variants identified in patients are sequenced in controls as previously described [12].
Bioinformatics analysis
Seattle SEQ annotation version 5.06 and dbSNP-Q were used for annotation purposes and queries on dbSNP, HapMap and 1000 Genomes project databases. Polyphen and SIFT (Sorting Intolerant From Tolerant; scores <0.05 = deleterious are considered deleterious) were used to assess the effect of non-synonymous/missense changes on the protein. Human Splicing Finder Version 2.4.1 (default) was used to investigate whether the nucleotide changes disrupted/created exonic splicing enhancers (ESEs) or silencers (ESS), or branch or splice-sites [24]. Default thresholds were used. To assess whether rare silent or missense variants could affect the secondary structure/stability of the RNA we used RNAmute. Variants to be submitted to RNAmute were selected according to 1) their level of conservation at both DNA and protein level and ii) by consulting/accessing the UCSC Evofold UCSC track (41way multi-alignment, hg19) [25]. Conservation was inferred from i) genomic base wise conservation scores calculated using PhyloP from the PHAST package (28-species multiple alignment) and from ii) the coded amino acid conservation examined using ConSurf web server. Among “silent” variants, those variants in the most conserved sites were thought to have potential functional relevance [26], [27].
Haplotype reconstruction
In order to phase rare variants present in individuals heterozygous for the RET risk haplotype (comprising the functional rs2435357 RET intron 1 SNP), we used the genotypes of 21 RET SNP which generated for 430 patients in our preceding studies. The RET rare variants identified were used as a SNP (one at the time) to re-construct haplotypes comprising the rare allele using the statistical software package PHASE version 2 as previously described [2], [4], [20]. A relation of the 21 SNPs can be found in the Cornes et al. manuscript [4].
Supporting Information
Figure S1.
Schematic representation of the impact of the variants Q327Q, L465L, L1077L, L651L, P841P, G954G F961L, Y1062C on RNA secondary structure (see main text and Material S1).
https://doi.org/10.1371/journal.pone.0028986.s001
(PDF)
Table S1.
A: distribution of the 61 rare RET variants across genders. B: characteristics of the HSCR patients with the R114H variant (N = 38).
https://doi.org/10.1371/journal.pone.0028986.s002
(DOCX)
Table S2.
Bioinformatics prediction of the rare RET variants identified.
https://doi.org/10.1371/journal.pone.0028986.s003
(DOCX)
Table S3.
Distribution of the 55 CDS rare variants in the RET protein domains (excludes intronic or 5′UT variants).
https://doi.org/10.1371/journal.pone.0028986.s004
(DOCX)
Table S4.
Rare RET variants identified in HSCR patients.
https://doi.org/10.1371/journal.pone.0028986.s005
(DOCX)
Author Contributions
Conceived and designed the experiments: M-MG-B PK-HT P-CS SSC. Performed the experiments: M-TS X-PM DNN LC TY-YL BKC GC CS-MT. Analyzed the data: M-TS X-PM DNN LC TY-YL BKC GC CS-MT. Contributed reagents/materials/analysis tools: ES-WN VC-HL X-ZW BW HW Z-WY L-MH LL HX DZ JL TLN IH-YC PH-YC X-LL RZ KK-YW. Wrote the paper: M-TS M-MG-B.
References
- 1. Amiel J, Sproat-Emison E, Garcia-Barcelo M, Lantieri F, Burzynski G, et al. (2008) Hirschsprung disease, associated syndromes and genetics: a review. J MedGenet 45: 1–14.
- 2. Emison ES, Garcia-Barcelo M, Grice EA, Lantieri F, Amiel J, et al. (2010) Differential contributions of rare and common, coding and noncoding ret mutations to multifactorial Hirschsprung disease liability. Am J Hum Genet 87: 60–74.
- 3. Manie S, Santoro M, Fusco A, Billaud M (2001) The RET receptor: function in development and dysfunction in congenital malformation. Trends Genet 17: 580–589.
- 4. Cornes BK, Tang CS, Leon TY, Hui KJ, So MT, et al. (2010) Haplotype analysis reveals a possible founder effect of RET mutation R114H for Hirschsprung's disease in the Chinese population. PLoS One 5: e10918.
- 5. Pelet A, Geneste O, Edery P, Pasini A, Chappuis S, et al. (1998) Various mechanisms cause RET-mediated signaling defects in Hirschsprung's disease. JClinInvest 101: 1415–1423.
- 6. Attie T, Pelet A, Edery P, Eng C, Mulligan LM, et al. (1995) Diversity of RET proto-oncogene mutations in familial and sporadic Hirschsprung disease. HumMolGenet 4: 1381–1386.
- 7. Edery P, Lyonnet S, Mulligan LM, Pelet A, Dow E, et al. (1994) Mutations of the RET proto-oncogene in Hirschsprung's disease. Nature 367: 378–380.
- 8. Seri M, Yin L, Barone V, Bolino A, Celli I, et al. (1997) Frequency of RET mutations in long- and short-segment Hirschsprung disease. HumMutat 9: 243–249.
- 9. Hofstra RM, Wu Y, Stulp RP, Elfferich P, Osinga J, et al. (2000) RET and GDNF gene scanning in Hirschsprung patients using two dual denaturing gel systems. HumMutat 15: 418–429.
- 10. Bolk S, Pelet A, Hofstra RM, Angrist M, Salomon R, et al. (2000) A human model for multigenic inheritance: phenotypic expression in Hirschsprung disease requires both the RET gene and a new 9q31 locus. ProcNatlAcadSciUSA 97: 268–273.
- 11. Gabriel SB, Salomon R, Pelet A, Angrist M, Amiel J, et al. (2002) Segregation at three loci explains familial and population risk in Hirschsprung disease. NatGenet 31: 89–93.
- 12. Garcia-Barcelo M, Sham MH, Lee WS, Lui VC, Chen BL, et al. (2004) Highly Recurrent RET Mutations and Novel Mutations in Genes of the Receptor Tyrosine Kinase and Endothelin Receptor B Pathways in Chinese Patients with Sporadic Hirschsprung Disease. ClinChem 50: 93–100.
- 13. Iwashita T, Kurokawa K, Qiao S, Murakami H, Asai N, et al. (2001) Functional analysis of RET with Hirschsprung mutations affecting its kinase domain. Gastroenterology 121: 24–33.
- 14. Iwashita T, Murakami H, Asai N, Takahashi M (1996) Mechanism of ret dysfunction by Hirschsprung mutations affecting its extracellular domain. HumMolGenet 5: 1577–1580.
- 15. Emison ES, McCallion AS, Kashuk CS, Bush RT, Grice E, et al. (2005) A common sex-dependent mutation in a RET enhancer underlies Hirschsprung disease risk. Nature 434: 857–863.
- 16. Garcia-Barcelo MM, Sham MH, Lui VC, Chen BL, Song YQ, et al. (2003) Chinese patients with sporadic Hirschsprung's disease are predominantly represented by a single RET haplotype. J MedGenet 40: e122.
- 17. Garcia-Barcelo M, Ganster RW, Lui VC, Leon TY, So MT, et al. (2005) TTF-1 and RET promoter SNPs: regulation of RET transcription in Hirschsprung's disease. HumMolGenet 14: 191–204.
- 18. Miao X, Leon TY, Ngan ES, So MT, Yuan ZW, et al. (2010) Reduced RET expression in gut tissue of individuals carrying risk alleles of Hirschsprung's disease. Hum Mol Genet 19: 1461–1467.
- 19. Dickson SP, Wang K, Krantz I, Hakonarson H, Goldstein DB (2010) Rare variants create synthetic genome-wide associations. PLoS Biol 8: e1000294.
- 20. Garcia-Barcelo MM, Tang CS, Ngan ES, Lui VC, Chen Y, et al. (2009) Genome-wide association study identifies NRG1 as a susceptibility locus for Hirschsprung's disease. ProcNatlAcadSciUSA 106: 2694–2699.
- 21. Won HH, Kim HJ, Lee KA, Kim JW (2008) Cataloging coding sequence variations in human genome databases. PLoS One 3: e3575.
- 22. Tarpey PS, Smith R, Pleasance E, Whibley A, Edkins S, et al. (2009) A systematic, large-scale resequencing screen of X-chromosome coding exons in mental retardation. Nat Genet 41: 535–543.
- 23. Sauna ZE, Kimchi-Sarfaty C (2011) Understanding the contribution of synonymous mutations to human disease. Nat Rev Genet 12: 683–691.
- 24. Desmet FO, Hamroun D, Lalande M, Collod-Beroud G, Claustres M, et al. (2009) Human Splicing Finder: an online bioinformatics tool to predict splicing signals. Nucleic Acids Res 37: e67.
- 25. Pedersen JS, Bejerano G, Siepel A, Rosenbloom K, Lindblad-Toh K, et al. (2006) Identification and classification of conserved RNA secondary structures in the human genome. PLoS Comput Biol 2: e33.
- 26. Pupko T, Bell RE, Mayrose I, Glaser F, Ben-Tal N (2002) Rate4Site: an algorithmic tool for the identification of functional regions in proteins by surface mapping of evolutionary determinants within their homologues. Bioinformatics 18(Suppl 1): S71–77.
- 27. Glaser F, Pupko T, Paz I, Bell RE, Bechor-Shental D, et al. (2003) ConSurf: identification of functional regions in proteins by surface-mapping of phylogenetic information. Bioinformatics 19: 163–164.