| Literature DB >> 19718455 |
Fabio Verginelli1, Faraz Bishehsari, Francesco Napolitano, Mahboobeh Mahdavinia, Alessandro Cama, Reza Malekzadeh, Gennaro Miele, Giancarlo Raiconi, Roberto Tagliaferri, Renato Mariani-Costantini.
Abstract
BACKGROUND:Entities:
Mesh:
Year: 2009 PMID: 19718455 PMCID: PMC2730577 DOI: 10.1371/journal.pone.0006824
Source DB: PubMed Journal: PLoS One ISSN: 1932-6203 Impact factor: 3.240
Figure 1Hierarchical clusterings for TP53 mutation sites, TP53 mutation types and food patterns.
Hierarchical clusterings (HC) by country/geographic area and silhouette plots for: A–B, TP53 mutation sites (MS); C–D, TP53 mutation types (MT); E–F, food patterns (FP).
Figure 2Mantel test correlation between TP53 mutation sites and TP53 mutation types.
Mantel test shows correlation between the distance matrices of mutation sites (MS) and mutation types (MT), with regression line parameter: r = 0.581, c = 0.015; R2 = 0.338, P = 0.001 after 10,000 permutations.
Figure 3Feature selection by heuristic approach for the mutation sites data.
Similarity values are on the y axis; number of features (i.e., codons) on the x axis, respectively ranked in decreasing order of weight (number of mutations, panel A), and decreasing order of the first 11 PC coefficients of each feature (where 11 was the number of PCs contributing 100% of the data variance, panel B). Panel C lists (in order of number) the codons (features) with highest variance in the mutation sites (MS) data, selected by weight ranking (23/173 codons) and by PC coefficients ranking (22/173 codons). Overall 19 codons, including the five TP53 mutation hotspots (highlighted in yellow), were selected by both methods.
Figure 4Scree and Kaiser's tests applied to the analysis of variance by PCA of the datamatrices for mutation sites, mutation types and food patterns.
Results obtained using Scree test are shown in panels A–C. Total mutation sites (MS) variance (panel A) is explained by 11 components, of which only 9 are visualized, being the Scree test cut at the 98% level. Four components contribute 80% of the MS variance, the first accounting for 31%. Total mutation types (MT) variance (B) is explained by 4 components, the first of which, by far the most relevant, contributes 65% of variance. Total food patterns (FP) variance (C) is explained by 3 components, the first of which contributes 87.3% of variance. Results obtained using Kaiser's test are shown in panels D–F. The first 11 PCs for mutation sites (MS) (A), the first 3 PCs for mutation types (MT) (B) and the first 4 PCs for food patterns (FP) (C) have eigenvalues above 1 (red line).
Worst-case exhaustive feature analysis of the mutation types datamatrix.
| |S| = number of selected features (j) | 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 |
| Feature (i) | ||||||||
|
| 0.42 | 0.39 | 0.32 | 0.38 | 0.31 | 0.38 | 0.38 | 1.00 |
|
| 0.49 | 0.44 | 0.34 | 0.39 | 0.31 | 0.39 | 0.38 | 1.00 |
|
| 0.44 | 0.40 | 0.32 | 0.40 | 0.38 | 0.38 | 0.38 | 1.00 |
|
| 0.50 | 0.40 | 0.34 | 0.33 | 0.31 | 0.38 | 0.38 | 1.00 |
|
| 0.42 | 0.40 | 0.37 | 0.33 | 0.38 | 0.38 | 0.38 | 1.00 |
|
|
|
|
|
|
|
|
|
|
|
| 0.36 | 0.37 | 0.32 | 0.33 | 0.31 | 0.38 | 0.38 | 1.00 |
|
| 0.51 | 0.37 | 0.37 | 0.33 | 0.31 | 0.38 | 0.38 | 1.00 |
Worst similarity values for each subset S of the MT features, including the selected feature in the left column, are computed using the exhaustive multivariate feature analysis. Bold characters highlight the worst similarity values of the features that most influence cluster structure. Entry (i,j) reports the minimum similarity value obtained using the i-th feature together with any other j-1 features. For example, feature A:T>C:G gives a similarity value that is at least 0.38 when coupled with any 6 of the features in the left column.
Best-case exhaustive feature analysis of the mutation types datamatrix.
| Column 1(Feature) | Column 2 (Features that paired with feature in column 1 give the same clustering obtained with all features) |
| A:T>C:G | G:C>A:T at CpG |
| A:T>G:C | none |
| A:T>T:A | none |
| FS | none |
| G:C>A:T | G:C>A:T at CpG |
| G:C>A:T at CpG | A:T>C:G and/or G:C>A:T and/or G:C>C:G |
| G:C>C:G | G:C>A:T at CpG |
| G:C>T:A | none |
Exhaustive multivariate feature analysis is used to highlight pairs of MT features giving similarity value equal to 1. Features in column 1 give similarity 1 if and only if coupled with one of the features in column 2. For example, feature A:T>C:G gives a similarity value 1 when coupled with G:C>A:T at CpGs. Relevance of individual features in column 1 is based on the number of other features in column 2 with which it can yield unitary value after pairing. The most relevant feature is G:C>A:T at CpGs, followed by A:T>C:G, G:C>A:T and G:C>C:G.
Figure 5Box-plots of feature relevance for mutation sites, mutation types and food patterns.
In each panel, box plots pertaining to clusters I versus II of mutation sites (MS), mutation types (MT) and food patterns (FP) obtained by hierarchical clustering are color-coded in red (cluster I) and blue (cluster II), respectively. Panel A: frequency box-plots of mutations at the 19 codons with highest weights and highest PCA variance coefficients (identified in Figure 4). Higher mutation frequencies at the three major CpG hotspot codons 175, 248, and 273 in I-MS versus II-MS are evident. Panel B: frequency box-plots of the 8 mutation types, showing higher frequencies of transitions at CpGs in I-MT (range: 46.1–61.2%) versus II-MT (range: 41.2–43.3%). Panel C: box-plots of the mean percent of the total available caloric value from each major food group in the relevant countries/geographic areas, showing lower cereals for the countries/geographic areas in I-FP versus those in II-FP, balanced by higher meat, milk, sweeteners and animal fats.
Figure 6Geographic visualization of the most relevant features evidenced in the box-plots.
Selected features of the mutation sites (MS, A), mutation types (MT, B) and food availability patterns (FP, C) datamatrices were mutations at the three major TP53 hotspot codons 175, 248, and 273 for MS; G:C>A:T mutations at CpGs for MT; meat/milk/sweeteners/animal fats (added), cereals (subtracted) for FP. Feature frequencies were summed and projected in yellow to red color range onto the geographic profiles of the relevant countries/geographic areas.
Worst-case exhaustive feature analysis of the food patterns datamatrix.
| |S| = number of selected features (j) | 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 10 | 11 | 12 | 13 |
| Feature (i) | |||||||||||||
|
| 0.53 | 0.43 | 0.43 | 0.45 | 0.54 | 0.54 | 0.54 | 0.54 | 0.54 | 0.71 | 1.00 | 1.00 | 1.00 |
|
| 1.00 | 0.51 | 0.51 | 0.51 | 0.51 | 0.53 | 0.53 | 0.54 | 0.54 | 0.71 | 1.00 | 1.00 | 1.00 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 0.47 | 0.34 | 0.32 | 0.36 | 0.40 | 0.53 | 0.53 | 0.54 | 0.54 | 0.71 | 1.00 | 1.00 | 1.00 |
|
| 0.65 | 0.52 | 0.52 | 0.51 | 0.49 | 0.53 | 0.53 | 0.54 | 0.54 | 0.71 | 1.00 | 1.00 | 1.00 |
|
| 0.71 | 0.71 | 0.71 | 0.71 | 0.71 | 0.71 | 0.72 | 0.72 | 0.72 | 0.80 | 1.00 | 1.00 | 1.00 |
|
| 1.00 | 0.80 | 0.71 | 0.71 | 0.71 | 0.71 | 0.71 | 0.71 | 0.71 | 0.71 | 1.00 | 1.00 | 1.00 |
|
| 0.55 | 0.37 | 0.39 | 0.36 | 0.40 | 0.53 | 0.53 | 0.54 | 0.54 | 0.71 | 1.00 | 1.00 | 1.00 |
|
| 0.40 | 0.40 | 0.35 | 0.36 | 0.40 | 0.53 | 0.53 | 0.54 | 0.54 | 0.71 | 1.00 | 1.00 | 1.00 |
|
| 0.40 | 0.31 | 0.32 | 0.36 | 0.40 | 0.53 | 0.53 | 0.54 | 0.54 | 0.71 | 1.00 | 1.00 | 1.00 |
|
| 0.58 | 0.56 | 0.54 | 0.54 | 0.54 | 0.54 | 0.54 | 0.54 | 0.54 | 0.71 | 1.00 | 1.00 | 1.00 |
|
| 0.51 | 0.51 | 0.51 | 0.51 | 0.51 | 0.54 | 0.54 | 0.70 | 0.70 | 1.00 | 1.00 | 1.00 | 1.00 |
|
| 0.35 | 0.31 | 0.32 | 0.49 | 0.40 | 0.53 | 0.53 | 0.54 | 0.54 | 0.71 | 1.00 | 1.00 | 1.00 |
Worst similarity values for each subset S of the FP features, including the selected feature in the left column, are computed using the exhaustive multivariate feature analysis. Bold characters highlight the worst similarity values of the features that most influence cluster structure (cereals). Entry (i,j) reports the minimum similarity value obtained using the i-th feature together with any other j-1 features. For example, feature “animal products” (i = 2) gives a similarity at least equal to 0.71 combined with any other remaining 9 features (j = 10).
Best-case exhaustive feature analysis of the food patterns datamatrix.
| Column 1 (Feature) | Column 2 (Features that paired with feature in column 1 give the same clustering obtained with all features) |
| Animal fats | Cereals, Meat, Milk |
| Animal products | Cereals, Meat, Oilcrops, Vegetable Oils |
| Cereals | All |
| Fish/Seafood | Cereals, Meat,Milk |
| Fruit | Cereals, Milk |
| Meat | Animal fats, Cereals, Fish/Seafood, Milk, Oilcrops, Starchy roots, Vegetable oils |
| Milk | All but Sweeteners |
| Oilcrops | Animal Products, Cereals, Meat, Milk |
| Pulses | Cereals, Milk |
| Starchy roots | Cereals, Meat, Milk |
| Sweeteners | Cereals |
| Vegetable oils | Cereals, Meat, Milk |
| Vegetables | Animal products, Cereals, Milk |
Exhaustive multivariate feature analysis is used to highlight pairs of FP features giving similarity value equal to 1. Features in column 1 give similarity equal to 1 if and only if coupled with one of the features in column 2. For example, feature cereals (i = 3) gives a similarity that is always equal to 1 alone or together with any set of other features. Relevance of individual features in column 1 is based on the number of other features in column 2 with which it can yield unitary value after pairing. The most relevant features are cereals, milk, and meat.
Pearson's correlation scores between the PCs of mutation types and food patterns.
| PCs | 1th PC (MT) | 2nd PC (MT) | 3rd PC (MT) |
|
|
| 0.2534 |
|
|
|
| −0.1735 | −0.4654 |
|
| 0.1724 |
| −0.1058 |
Pearson's correlation scores computed between the principal components (PC) of mutation types (MT) and food availability patterns (FP), with the corresponding P-values (square brackets). Significant correlations are highlighted in bold and the corresponding r-coefficient values are marked with an asterisk (*).
Pearson's correlation scores between the PCs of mutation types and mutation sites.
| PCs | 1th PC (MT) | 2nd PC (MT) | 3rd PC (MT) |
|
|
| −0.0908 | −0.0118 |
|
| 0.0975 | −0.1414 | −0.0568 |
|
| −0.0069 | 0.2340 | 0.3828 |
Pearson's correlation scores computed between the principal components (PC) of mutation types (MT) and mutation sites (MS), with the corresponding P-values (square brackets). Significant correlations are highlighted in bold and the corresponding r-coefficient values are marked with an asterisk (*).
Pearson's correlation scores between the PCs of mutation sites and food patterns.
| PCs | 1th PC (MS) | 2nd PC (MS) | 3rd PC (MS) |
|
| 0.4176 [ | −0.4246 [ | 0.4168 [ |
|
| 0.3858 [ | 0.1226 [ | −0.2474 [ |
|
| −0.2826 [ | −0.4014 [ | −0.1824 [ |
Pearson's correlation scores computed between the principal components (PC) of mutation sites (MS) and food availability patterns (FP), with the corresponding P-values (square brackets). Significant correlations are highlighted in bold and the corresponding r-coefficient values are marked with an asterisk (*).
Figure 7Scatter plot of the correlation between mutation sites and types according to countries/geographic areas.
Scatter plot of the countries/geographic areas in the TP53 database projected on the 2-dimensional space spanned by the first principal components of mutation sites (Mutation Site PC2) and mutation types (Mutation Type PC1). Italy, Iran, South and East Asia and West and Central Europe have relatively lower frequencies of mutations at CpG hotspot codons, compensated by higher frequencies of mutations at all other sites (see also box-plots in Figure 5). Mutation frequencies at CpG hotspots increase in other countries, with highest frequencies in Australia and UK. A linear regression shows the global trend of the correlation (r = −0.8742).
Figure 8Scatter plot of the correlation between mutation types and food patterns according to countries/geographic areas.
Scatter plot of the countries/geographic areas in the TP53 database projected on the 2-dimensional space spanned by the highest coefficient features of the first principal component (PC) of food availability pattern (FP), i.e., cereals, meat, milk, sweeteners, animal fats, and of the first PC of mutation type (MT), i.e., G:C>A:T at CpGs. A linear regression shows the global correlation trend.