Data import

db <- read_excel(here::here("data-raw/Dati per statistica FL.xlsx")) |> 
  
  janitor::clean_names() |> 
  
  mutate(across(c(group_1_covid_2_transplant_3_donatore,
                   sex_1_m_2_f,
                   status_vivo_1_morto_0,
                   ecmo_y_1_n_2,
                   comorbidities_1_y_1_n_0,
                   fibrin_y_1_n_2:fungal_infections_y_1_n_2), as.factor))

# dataReporter::makeDataReport(db,codebook = TRUE)

Infezione

• Vi sono differenze di infezioni tra i tre gruppi ?

Ho creato una nuova varaibile che si chiama inf_score che assume i valore 1 se c’è almeno una infezione, e 2 o 3 se ce ne sono più di una. Ovviamente 0 se non ce ne sono in quel paziente.

db_inf <- db |> select(group_1_covid_2_transplant_3_donatore, 
                       bacterial_infections_y_1_n_2,
                       viral_infections_y_1_n_2,
                       fungal_infections_y_1_n_2) |> 
  
  mutate(group_1_covid_2_transplant_3_donatore = case_when(
    group_1_covid_2_transplant_3_donatore == 1 ~ 'COVID',
    group_1_covid_2_transplant_3_donatore == 2 ~ 'Transplant',
    TRUE ~ 'Donor')) |> 
  
  mutate(inf_score = case_when(
    bacterial_infections_y_1_n_2 == 1 & viral_infections_y_1_n_2 == 1 & fungal_infections_y_1_n_2 == 1 ~ 3,
    
    
    bacterial_infections_y_1_n_2 == 1 & viral_infections_y_1_n_2 == 1 ~ 2,
    bacterial_infections_y_1_n_2 == 1 & fungal_infections_y_1_n_2 == 1 ~ 2,
    viral_infections_y_1_n_2 == 1 & fungal_infections_y_1_n_2 == 1 ~ 2,
    
    bacterial_infections_y_1_n_2 == 1 ~ 1,
    viral_infections_y_1_n_2 == 1 ~ 1,
    fungal_infections_y_1_n_2 == 1 ~ 1,
    TRUE ~ 0
  ))
t0 <- db_inf |>  
  
  tbl_summary(by = group_1_covid_2_transplant_3_donatore)

t1 <- db_inf |> 
  filter(group_1_covid_2_transplant_3_donatore %in% c("COVID", "Transplant")) |> 
  
  tbl_summary(by = group_1_covid_2_transplant_3_donatore) |> 
  
  add_p() |> 
  
  modify_header(p.value ~ gt::md("**COVID vs. Transplant**")) |> 
  # hide summary stat columns
  modify_table_styling(all_stat_cols(), hide = TRUE)
  
t2 <- db_inf |> 
  filter(group_1_covid_2_transplant_3_donatore %in% c("COVID", "Donor")) |> 
  
  tbl_summary(by = group_1_covid_2_transplant_3_donatore) |> 
  
  add_p() |> 
  
  modify_header(p.value ~ gt::md("**COVID vs. Donor**")) |> 
  # hide summary stat columns
  modify_table_styling(all_stat_cols(), hide = TRUE)

t3 <- db_inf |> 
  filter(group_1_covid_2_transplant_3_donatore %in% c("Transplant", "Donor")) |> 
  
  tbl_summary(by = group_1_covid_2_transplant_3_donatore) |> 
  
  add_p() |> 
  
  modify_header(p.value ~ gt::md("**Transplant vs. Donor**")) |> 
  # hide summary stat columns
  modify_table_styling(all_stat_cols(), hide = TRUE)

t_merge <- tbl_merge(list(t0, t1, t2, t3)) %>%
    modify_spanning_header(
      list(
        all_stat_cols() ~ "**Groups**",
        starts_with("p.value") ~ "**P-values**") 
    )
t_merge
Characteristic Groups P-values
COVID, N = 161 Donor, N = 121 Transplant, N = 121 COVID vs. Transplant2 COVID vs. Donor2 Transplant vs. Donor2
bacterial_infections_y_1_n_2 0.7 0.053 0.037
1 5 (31%) 0 (0%) 5 (42%)
2 11 (69%) 12 (100%) 7 (58%)
viral_infections_y_1_n_2 0.7 0.053 0.037
1 5 (31%) 0 (0%) 5 (42%)
2 11 (69%) 12 (100%) 7 (58%)
fungal_infections_y_1_n_2 0.053 0.053 >0.9
1 5 (31%) 0 (0%) 0 (0%)
2 11 (69%) 12 (100%) 12 (100%)
inf_score >0.9 <0.001 <0.001
0 5 (31%) 12 (100%) 3 (25%)
1 8 (50%) 0 (0%) 8 (67%)
2 2 (12%) 0 (0%) 1 (8.3%)
3 1 (6.2%) 0 (0%) 0 (0%)
1 n (%)
2 Fisher's exact test

• L’associaz con gli i ag infettivi era maggiormente presente in paz COVID con lunga data di malattia

Associazione di infezioni (si/no) con durata della malattia, i.e. gli infetti sono hanno ICU LOS o H LOS più lunghe?

• La maggiore neutrofilia nei COVID era correlata con le infezioni o era presente a prescindere ?

Solo COVID

db_los <- db |> select(
  
  icu_los, 
  h_los, 
  bacterial_infections_y_1_n_2,
  viral_infections_y_1_n_2,
  fungal_infections_y_1_n_2,
  group_1_covid_2_transplant_3_donatore,
  neutrophilic_granulocytes_percent
  
) |> 
  
  mutate(across(c(bacterial_infections_y_1_n_2:fungal_infections_y_1_n_2), ~ifelse(.==1, 'Yes', 'No')))





tbl_summary(db_los |> select(icu_los, 
                             h_los, 
                             bacterial_infections_y_1_n_2, 
                             group_1_covid_2_transplant_3_donatore,
                             neutrophilic_granulocytes_percent) |> 
              
              filter(group_1_covid_2_transplant_3_donatore == '1'), 
            
            by = bacterial_infections_y_1_n_2,
            missing = 'no', 
            type = c(neutrophilic_granulocytes_percent) ~ 'continuous'
            ) |> 
  
  add_p() |> 
  
  modify_caption('**Bacterial**')
## Warning for variable 'icu_los':
## simpleWarning in wilcox.test.default(x = DATA[[1L]], y = DATA[[2L]], ...): cannot compute exact p-value with ties
## Warning for variable 'h_los':
## simpleWarning in wilcox.test.default(x = DATA[[1L]], y = DATA[[2L]], ...): cannot compute exact p-value with ties
## Warning for variable 'neutrophilic_granulocytes_percent':
## simpleWarning in wilcox.test.default(x = DATA[[1L]], y = DATA[[2L]], ...): cannot compute exact p-value with ties
Bacterial
Characteristic No, N = 111 Yes, N = 51 p-value2
icu_los 24 (16, 40) 15 (15, 18) 0.4
h_los 33 (30, 60) 30 (18, 60) 0.5
group_1_covid_2_transplant_3_donatore >0.9
1 11 (100%) 5 (100%)
2 0 (0%) 0 (0%)
3 0 (0%) 0 (0%)
neutrophilic_granulocytes_percent 60 (45, 85) 50 (40, 90) >0.9
1 Median (IQR); n (%)
2 Wilcoxon rank sum test; Fisher's exact test 
tbl_summary(db_los |> select(icu_los, 
                             h_los, 
                             viral_infections_y_1_n_2, 
                             group_1_covid_2_transplant_3_donatore,
                             neutrophilic_granulocytes_percent) |> 
              
              filter(group_1_covid_2_transplant_3_donatore == '1'), 
            by = viral_infections_y_1_n_2,
            missing = 'no', 
            type = c(neutrophilic_granulocytes_percent) ~ 'continuous') |> 
  
  add_p() |> 
  
  modify_caption('**Viral**')
## Warning for variable 'icu_los':
## simpleWarning in wilcox.test.default(x = DATA[[1L]], y = DATA[[2L]], ...): cannot compute exact p-value with ties
## Warning for variable 'h_los':
## simpleWarning in wilcox.test.default(x = DATA[[1L]], y = DATA[[2L]], ...): cannot compute exact p-value with ties
## Warning for variable 'neutrophilic_granulocytes_percent':
## simpleWarning in wilcox.test.default(x = DATA[[1L]], y = DATA[[2L]], ...): cannot compute exact p-value with ties
Viral
Characteristic No, N = 111 Yes, N = 51 p-value2
icu_los 23 (12, 30) 18 (18, 24) 0.7
h_los 30 (28, 60) 33 (30, 41) 0.8
group_1_covid_2_transplant_3_donatore >0.9
1 11 (100%) 5 (100%)
2 0 (0%) 0 (0%)
3 0 (0%) 0 (0%)
neutrophilic_granulocytes_percent 50 (40, 85) 60 (50, 90) 0.6
1 Median (IQR); n (%)
2 Wilcoxon rank sum test; Fisher's exact test 
tbl_summary(db_los |> select(icu_los, 
                             h_los, 
                             fungal_infections_y_1_n_2, 
                             group_1_covid_2_transplant_3_donatore,
                             neutrophilic_granulocytes_percent) |> 
              
              filter(group_1_covid_2_transplant_3_donatore == '1'),
            by = fungal_infections_y_1_n_2,
            missing = 'no', 
            type = c(neutrophilic_granulocytes_percent) ~ 'continuous') |> 
  
  add_p() |> 
  
  modify_caption('**Fungal**')
## Warning for variable 'icu_los':
## simpleWarning in wilcox.test.default(x = DATA[[1L]], y = DATA[[2L]], ...): cannot compute exact p-value with ties
## Warning for variable 'h_los':
## simpleWarning in wilcox.test.default(x = DATA[[1L]], y = DATA[[2L]], ...): cannot compute exact p-value with ties
## Warning for variable 'neutrophilic_granulocytes_percent':
## simpleWarning in wilcox.test.default(x = DATA[[1L]], y = DATA[[2L]], ...): cannot compute exact p-value with ties
Fungal
Characteristic No, N = 111 Yes, N = 51 p-value2
icu_los 18 (12, 24) 28 (26, 53) 0.079
h_los 33 (22, 50) 30 (30, 79) 0.5
group_1_covid_2_transplant_3_donatore >0.9
1 11 (100%) 5 (100%)
2 0 (0%) 0 (0%)
3 0 (0%) 0 (0%)
neutrophilic_granulocytes_percent 50 (40, 85) 60 (50, 90) 0.3
1 Median (IQR); n (%)
2 Wilcoxon rank sum test; Fisher's exact test

C’è una tendenza per le infezioni fungine, non per il resto.

Cellulare

• Vi sono differenze di cellularità (macrofagi, neutrofili, linfociti, fibrina, sangue e pneumociti reattivi tra i due gruppi (COVID vs IC controls) ?

db_cel <- db |> select(group_1_covid_2_transplant_3_donatore,
                       macrophages_percent:hyperplastic_pneumocytes_y_1_n_2) |> 
  
  filter(group_1_covid_2_transplant_3_donatore == 1 | group_1_covid_2_transplant_3_donatore == 2) |> 
  
  mutate(group_1_covid_2_transplant_3_donatore = case_when(
    group_1_covid_2_transplant_3_donatore == 1 ~ 'COVID',
    TRUE ~ 'Transplant')) |> 
  
  droplevels() # altrimenti tieni anche i donatori anche se non ci sono più




tbl_summary(db_cel, by = group_1_covid_2_transplant_3_donatore,
            type = c(lymphocytes_percent) ~ 'continuous') |> 
  add_p()
## Warning for variable 'macrophages_percent':
## simpleWarning in wilcox.test.default(x = DATA[[1L]], y = DATA[[2L]], ...): cannot compute exact p-value with ties
## Warning for variable 'lymphocytes_percent':
## simpleWarning in wilcox.test.default(x = DATA[[1L]], y = DATA[[2L]], ...): cannot compute exact p-value with ties
## Warning for variable 'neutrophilic_granulocytes_percent':
## simpleWarning in wilcox.test.default(x = DATA[[1L]], y = DATA[[2L]], ...): cannot compute exact p-value with ties
Characteristic COVID, N = 161 Transplant, N = 121 p-value2
macrophages_percent 40 (10, 52) 90 (90, 95) <0.001
lymphocytes_percent 0.0 (0.0, 0.0) 6.5 (5.0, 10.0) <0.001
neutrophilic_granulocytes_percent 55 (40, 90) 0 (0, 2) <0.001
fibrin_y_1_n_2 <0.001
1 16 (100%) 0 (0%)
2 0 (0%) 12 (100%)
eritrocytes_y_1_n_2 <0.001
1 15 (94%) 3 (25%)
2 1 (6.2%) 9 (75%)
hyperplastic_pneumocytes_y_1_n_2 0.001
1 14 (88%) 3 (25%)
2 2 (12%) 9 (75%)
1 Median (IQR); n (%)
2 Wilcoxon rank sum test; Pearson's Chi-squared test; Fisher's exact test

E’ tutto differente

Immunologia

tbl_summary(db |> select(group_1_covid_2_transplant_3_donatore, 
                         ifna1_delta_ct:tnf_delta_ct
                         ),
            by = group_1_covid_2_transplant_3_donatore,
            
            type = everything() ~ "continuous") |> 
  
  add_p()
Characteristic 1, N = 161 2, N = 121 3, N = 121 p-value2
ifna1_delta_ct 7 (6, 8) 9 (7, 11) 7 (6, 7) 0.065
ifna16_delta_ct 12 (10, 30) 30 (26, 30) 11 (11, 30) 0.2
ifna17_delta_ct 8.7 (7.9, 10.6) 9.4 (8.4, 10.6) 8.2 (7.4, 9.0) 0.13
ifna2_delta_ct 9 (8, 10) 9 (9, 11) 8 (8, 9) 0.2
ifna6_delta_ct 7.7 (7.0, 9.0) 8.2 (7.4, 9.2) 7.0 (6.6, 7.5) 0.2
ifna7_delta_ct 7.21 (6.86, 9.03) 9.35 (8.14, 10.00) 7.19 (6.89, 7.82) 0.024
ifna8_delta_ct 9 (8, 10) 9 (8, 10) 8 (7, 8) 0.038
ifnb1_delta_ct 8 (8, 10) 8 (8, 10) 8 (7, 9) 0.3
ifng_delta_ct 30 (30, 30) 30 (-11, 30) 30 (30, 30) 0.041
il10_delta_ct 30 (30, 30) 30 (30, 30) 30 (26, 30) 0.2
il12a_delta_ct 30 (30, 30) 30 (30, 30) 30 (30, 30) 0.7
il12b_delta_ct 30 (21, 30) 30 (30, 30) 30 (30, 30) 0.4
il13_delta_ct 30 (30, 30) 30 (30, 30) 30 (30, 30) 0.094
il15_delta_ct 30 (30, 30) 30 (30, 30) 30 (30, 30) >0.9
il16_delta_ct 30 (30, 30) 22 (12, 30) 30 (30, 30) 0.3
il17a_delta_ct 30 (30, 30) 30 (30, 30) 30 (30, 30) 0.3
il18_delta_ct 30 (30, 30) 30 (26, 30) 30 (30, 30) 0.2
il1a_delta_ct 30 (25, 30) 30 (12, 30) 30 (30, 30) 0.068
il1b_delta_ct 10 (8, 30) 12 (11, 30) 30 (30, 30) 0.005
il2_delta_ct 30 (30, 30) 30 (30, 30) 30 (30, 30) 0.5
il3_delta_ct 30.0000 (30.0000, 30.0000) 30.0000 (30.0000, 30.0000) 30.0000 (30.0000, 30.0000) 0.5
il4_delta_ct 30 (30, 30) 30 (30, 30) 30 (30, 30) 0.2
il5_delta_ct 30 (30, 30) 30 (30, 30) 30 (30, 30) 0.7
il6_delta_ct 30 (25, 30) 30 (26, 30) 30 (10, 30) 0.6
il8_delta_ct 21 (8, 30) 12 (10, 30) 30 (13, 30) 0.3
il9_delta_ct 12 (10, 30) 30 (30, 30) 30 (14, 30) 0.014
lta_delta_ct 30 (30, 30) 30 (30, 30) 30 (30, 30) 0.3
tnf_delta_ct 30 (16, 30) 30 (11, 30) 30 (30, 30) 0.2
1 Median (IQR)
2 Kruskal-Wallis rank sum test

IFNA1 e IL1A non sono più significative (lo erano senza gli ultimi 4 aggiunti). Metto lo stesso i grafici

# graphs of significants --------------------------------------------------

out <- c('ifna1_delta_ct', 'ifna7_delta_ct', 'ifna8_delta_ct', 'ifng_delta_ct', 'il1a_delta_ct', 'il1b_delta_ct', 'il9_delta_ct')

graphs<- function(out){
  
  ggbetweenstats(
    
    data = db,
    x = group_1_covid_2_transplant_3_donatore,
    y = {{out}},
    type = 'np',
    title = paste0(out)
  )
  
  
}

map(out, graphs)
## [[1]]

## 
## [[2]]

## 
## [[3]]

## 
## [[4]]

## 
## [[5]]

## 
## [[6]]

## 
## [[7]]

L’associazione più interessante sembre sia con IL9, che è completamente diversa nei COVID e quindi sembra “tipica” della malattia

• (solo COVID) Le citochine alterate erano maggiormente o meno alterate nei pazienti COVID con agenti infettivi ( eventualmente differenziando per agente microbico)

db_cito <- db |> select(
  
  bacterial_infections_y_1_n_2,
  viral_infections_y_1_n_2,
  fungal_infections_y_1_n_2,
  group_1_covid_2_transplant_3_donatore,
  ifna1_delta_ct, 
  ifna7_delta_ct, 
  ifna8_delta_ct, 
  ifng_delta_ct, 
  il1a_delta_ct, 
  il1b_delta_ct, 
  il9_delta_ct, 
  bmi,
  age_years
  
  
) |> 
  
  mutate(across(c(bacterial_infections_y_1_n_2:fungal_infections_y_1_n_2), ~ifelse(.==1, 'Yes', 'No')))





tbl_summary(db_cito |> select(-viral_infections_y_1_n_2, -fungal_infections_y_1_n_2) |> 
              
              filter(group_1_covid_2_transplant_3_donatore == '1'), 
            
            by = bacterial_infections_y_1_n_2,
            missing = 'no',
            type = c(ifng_delta_ct, il1a_delta_ct) ~ 'continuous'
            ) |> 
  
  add_p() |> 
  
  modify_caption('**Bacterial**')
## Warning for variable 'ifna1_delta_ct':
## simpleWarning in wilcox.test.default(x = DATA[[1L]], y = DATA[[2L]], ...): cannot compute exact p-value with ties
## Warning for variable 'ifna7_delta_ct':
## simpleWarning in wilcox.test.default(x = DATA[[1L]], y = DATA[[2L]], ...): cannot compute exact p-value with ties
## Warning for variable 'ifna8_delta_ct':
## simpleWarning in wilcox.test.default(x = DATA[[1L]], y = DATA[[2L]], ...): cannot compute exact p-value with ties
## Warning for variable 'ifng_delta_ct':
## simpleWarning in wilcox.test.default(x = DATA[[1L]], y = DATA[[2L]], ...): cannot compute exact p-value with ties
## Warning for variable 'il1a_delta_ct':
## simpleWarning in wilcox.test.default(x = DATA[[1L]], y = DATA[[2L]], ...): cannot compute exact p-value with ties
## Warning for variable 'il1b_delta_ct':
## simpleWarning in wilcox.test.default(x = DATA[[1L]], y = DATA[[2L]], ...): cannot compute exact p-value with ties
## Warning for variable 'il9_delta_ct':
## simpleWarning in wilcox.test.default(x = DATA[[1L]], y = DATA[[2L]], ...): cannot compute exact p-value with ties
## Warning for variable 'age_years':
## simpleWarning in wilcox.test.default(x = DATA[[1L]], y = DATA[[2L]], ...): cannot compute exact p-value with ties
Bacterial
Characteristic No, N = 111 Yes, N = 51 p-value2
group_1_covid_2_transplant_3_donatore >0.9
1 11 (100%) 5 (100%)
2 0 (0%) 0 (0%)
3 0 (0%) 0 (0%)
ifna1_delta_ct 7 (6, 8) 7 (5, 7) 0.6
ifna7_delta_ct 7 (7, 9) 7 (7, 7) 0.5
ifna8_delta_ct 9 (8, 10) 8 (8, 8) 0.11
ifng_delta_ct 30.0 (30.0, 30.0) 30.0 (30.0, 30.0) 0.2
il1a_delta_ct 30 (30, 30) 30 (8, 30) 0.3
il1b_delta_ct 10 (9, 30) 7 (6, 30) 0.5
il9_delta_ct 11 (9, 23) 14 (13, 30) 0.2
bmi 29 (26, 36) 38 (32, 40) 0.4
age_years 61 (57, 70) 50 (46, 73) 0.4
1 n (%); Median (IQR)
2 Fisher's exact test; Wilcoxon rank sum test; Wilcoxon rank sum exact test 
tbl_summary(db_cito |> select(-bacterial_infections_y_1_n_2, -fungal_infections_y_1_n_2) |> 
              
              filter(group_1_covid_2_transplant_3_donatore == '1'), 
            
            by = viral_infections_y_1_n_2,
            missing = 'no',
            type = c(ifng_delta_ct, il1a_delta_ct) ~ 'continuous'
            ) |> 
  
  add_p() |> 
  
  modify_caption('**Viral**')
## Warning for variable 'ifna1_delta_ct':
## simpleWarning in wilcox.test.default(x = DATA[[1L]], y = DATA[[2L]], ...): cannot compute exact p-value with ties
## Warning for variable 'ifna7_delta_ct':
## simpleWarning in wilcox.test.default(x = DATA[[1L]], y = DATA[[2L]], ...): cannot compute exact p-value with ties
## Warning for variable 'ifna8_delta_ct':
## simpleWarning in wilcox.test.default(x = DATA[[1L]], y = DATA[[2L]], ...): cannot compute exact p-value with ties
## Warning for variable 'ifng_delta_ct':
## simpleWarning in wilcox.test.default(x = DATA[[1L]], y = DATA[[2L]], ...): cannot compute exact p-value with ties
## Warning for variable 'il1a_delta_ct':
## simpleWarning in wilcox.test.default(x = DATA[[1L]], y = DATA[[2L]], ...): cannot compute exact p-value with ties
## Warning for variable 'il1b_delta_ct':
## simpleWarning in wilcox.test.default(x = DATA[[1L]], y = DATA[[2L]], ...): cannot compute exact p-value with ties
## Warning for variable 'il9_delta_ct':
## simpleWarning in wilcox.test.default(x = DATA[[1L]], y = DATA[[2L]], ...): cannot compute exact p-value with ties
## Warning for variable 'age_years':
## simpleWarning in wilcox.test.default(x = DATA[[1L]], y = DATA[[2L]], ...): cannot compute exact p-value with ties
Viral
Characteristic No, N = 111 Yes, N = 51 p-value2
group_1_covid_2_transplant_3_donatore >0.9
1 11 (100%) 5 (100%)
2 0 (0%) 0 (0%)
3 0 (0%) 0 (0%)
ifna1_delta_ct 7 (6, 19) 7 (6, 7) 0.5
ifna7_delta_ct 7 (7, 9) 7 (7, 7) >0.9
ifna8_delta_ct 9 (8, 20) 8 (8, 8) 0.4
ifng_delta_ct 30.0 (30.0, 30.0) 30.0 (30.0, 30.0) 0.6
il1a_delta_ct 30 (9, 30) 30 (30, 30) 0.2
il1b_delta_ct 9 (7, 22) 30 (10, 30) 0.15
il9_delta_ct 15 (10, 30) 11 (11, 13) 0.6
bmi 30 (28, 35) 40 (25, 50) 0.7
age_years 60 (51, 73) 61 (60, 64) >0.9
1 n (%); Median (IQR)
2 Fisher's exact test; Wilcoxon rank sum test; Wilcoxon rank sum exact test 
tbl_summary(db_cito |> select(-viral_infections_y_1_n_2, -bacterial_infections_y_1_n_2) |> 
              
              filter(group_1_covid_2_transplant_3_donatore == '1'), 
            
            by = fungal_infections_y_1_n_2,
            missing = 'no',
            type = c(ifng_delta_ct, il1a_delta_ct) ~ 'continuous'
            ) |> 
  
  add_p() |> 
  
  modify_caption('**Fungal**')
## Warning for variable 'ifna1_delta_ct':
## simpleWarning in wilcox.test.default(x = DATA[[1L]], y = DATA[[2L]], ...): cannot compute exact p-value with ties
## Warning for variable 'ifna7_delta_ct':
## simpleWarning in wilcox.test.default(x = DATA[[1L]], y = DATA[[2L]], ...): cannot compute exact p-value with ties
## Warning for variable 'ifna8_delta_ct':
## simpleWarning in wilcox.test.default(x = DATA[[1L]], y = DATA[[2L]], ...): cannot compute exact p-value with ties
## Warning for variable 'ifng_delta_ct':
## simpleWarning in wilcox.test.default(x = DATA[[1L]], y = DATA[[2L]], ...): cannot compute exact p-value with ties
## Warning for variable 'il1a_delta_ct':
## simpleWarning in wilcox.test.default(x = DATA[[1L]], y = DATA[[2L]], ...): cannot compute exact p-value with ties
## Warning for variable 'il1b_delta_ct':
## simpleWarning in wilcox.test.default(x = DATA[[1L]], y = DATA[[2L]], ...): cannot compute exact p-value with ties
## Warning for variable 'il9_delta_ct':
## simpleWarning in wilcox.test.default(x = DATA[[1L]], y = DATA[[2L]], ...): cannot compute exact p-value with ties
## Warning for variable 'age_years':
## simpleWarning in wilcox.test.default(x = DATA[[1L]], y = DATA[[2L]], ...): cannot compute exact p-value with ties
Fungal
Characteristic No, N = 111 Yes, N = 51 p-value2
group_1_covid_2_transplant_3_donatore >0.9
1 11 (100%) 5 (100%)
2 0 (0%) 0 (0%)
3 0 (0%) 0 (0%)
ifna1_delta_ct 7 (6, 19) 6 (5, 7) 0.2
ifna7_delta_ct 7 (7, 9) 7 (7, 8) 0.8
ifna8_delta_ct 8 (8, 20) 9 (8, 9) 0.6
ifng_delta_ct 30.0 (30.0, 30.0) 30.0 (30.0, 30.0) 0.6
il1a_delta_ct 30 (20, 30) 30 (30, 30) 0.9
il1b_delta_ct 10 (7, 30) 15 (10, 30) 0.5
il9_delta_ct 13 (10, 30) 11 (8, 14) 0.3
bmi 30 (27, 39) 31 (27, 50) 0.5
age_years 64 (53, 73) 60 (45, 60) 0.14
1 n (%); Median (IQR)
2 Fisher's exact test; Wilcoxon rank sum test; Wilcoxon rank sum exact test

No, nessuna differenza di espressione delle citochine in infetti e non (gruppo COVID)

Classificazione

• Relativamente ai pazienti COVID: Quale fattore era maggiormente predittivo di cattiva outcome: partern citochinico; coinfezioni ? caratteristiche del paz , pattern di citologiase puo applicare qualche algoritmo di machine learning sarebbe carino )

Gli outcome sono:  Mortalità  Long stay ICU  Utilizzo di ECMO  Ospedalizzazione

db_boruta <- db |> filter(group_1_covid_2_transplant_3_donatore ==1) |> 
  droplevels() |> 
  select(sex_1_m_2_f,
         age_years,
         bmi,
         comorbidities_1_y_1_n_0:tnf_delta_ct,
         
         status_vivo_1_morto_0,
         icu_los,
         h_los,
         ecmo_y_1_n_2)

Mortalità

# Boruta non vuole missing, li imputo sempre con RF
db_imp <- randomForestSRC::impute(data = as.data.frame(db_boruta))

# seed <- c(1, 23, 3011, 12345, 654321) #ne laascio solo uno perché sono consistenti

seed <- c(1, 23, 3011, 12345, 654321)

# creo una funzione per far girare boruta in automatico con tutti i seed
boruta_fun <- function(seed){
  set.seed(seed)
  
  boruta.lungs_train <- Boruta(status_vivo_1_morto_0 ~ . , 
         maxRuns = 1000, # aumento il numero per includere le variabili escluse
         data = db_imp |> select(-icu_los, -h_los, -ecmo_y_1_n_2))
  print(boruta.lungs_train)
  
  boruta.lungs <- TentativeRoughFix(boruta.lungs_train)
  print(boruta.lungs)
  
  plot <- plot(boruta.lungs, xlab = "", las = 3, cex.axis = 0.8, 
     whichShadow = c(FALSE, FALSE, FALSE),
     pars = par(mar = c(9.5,4,2,1)))

  lungs_df <- attStats(boruta.lungs) %>%
    relocate(decision) %>%
    dplyr::select(-normHits) %>% # così ci sta nella pagina
    arrange(decision, desc(meanImp))

  return(list(lungs_df, plot))
  
  
}

boruta_res <- map(seed, ~ boruta_fun(.x))
## Boruta performed 999 iterations in 30.05836 secs.
##  1 attributes confirmed important: age_years;
##  41 attributes confirmed unimportant: bacterial_infections_y_1_n_2,
## bmi, comorbidities_1_y_1_n_0, eritrocytes_y_1_n_2, fibrin_y_1_n_2 and
## 36 more;
##  1 tentative attributes left: tnf_delta_ct;
## Boruta performed 999 iterations in 30.05836 secs.
## Tentatives roughfixed over the last 999 iterations.
##  1 attributes confirmed important: age_years;
##  42 attributes confirmed unimportant: bacterial_infections_y_1_n_2,
## bmi, comorbidities_1_y_1_n_0, eritrocytes_y_1_n_2, fibrin_y_1_n_2 and
## 37 more;

## Boruta performed 999 iterations in 20.48649 secs.
##  1 attributes confirmed important: age_years;
##  41 attributes confirmed unimportant: bacterial_infections_y_1_n_2,
## bmi, comorbidities_1_y_1_n_0, eritrocytes_y_1_n_2, fibrin_y_1_n_2 and
## 36 more;
##  1 tentative attributes left: tnf_delta_ct;
## Boruta performed 999 iterations in 20.48649 secs.
## Tentatives roughfixed over the last 999 iterations.
##  1 attributes confirmed important: age_years;
##  42 attributes confirmed unimportant: bacterial_infections_y_1_n_2,
## bmi, comorbidities_1_y_1_n_0, eritrocytes_y_1_n_2, fibrin_y_1_n_2 and
## 37 more;
## Boruta performed 112 iterations in 2.664259 secs.
##  1 attributes confirmed important: age_years;
##  42 attributes confirmed unimportant: bacterial_infections_y_1_n_2,
## bmi, comorbidities_1_y_1_n_0, eritrocytes_y_1_n_2, fibrin_y_1_n_2 and
## 37 more;
## Warning in TentativeRoughFix(boruta.lungs_train): There are no Tentative
## attributes! Returning original object.

## Boruta performed 112 iterations in 2.664259 secs.
##  1 attributes confirmed important: age_years;
##  42 attributes confirmed unimportant: bacterial_infections_y_1_n_2,
## bmi, comorbidities_1_y_1_n_0, eritrocytes_y_1_n_2, fibrin_y_1_n_2 and
## 37 more;

## Boruta performed 999 iterations in 19.84405 secs.
##  1 attributes confirmed important: age_years;
##  41 attributes confirmed unimportant: bacterial_infections_y_1_n_2,
## bmi, comorbidities_1_y_1_n_0, eritrocytes_y_1_n_2, fibrin_y_1_n_2 and
## 36 more;
##  1 tentative attributes left: tnf_delta_ct;
## Boruta performed 999 iterations in 19.84405 secs.
## Tentatives roughfixed over the last 999 iterations.
##  1 attributes confirmed important: age_years;
##  42 attributes confirmed unimportant: bacterial_infections_y_1_n_2,
## bmi, comorbidities_1_y_1_n_0, eritrocytes_y_1_n_2, fibrin_y_1_n_2 and
## 37 more;
## Boruta performed 162 iterations in 2.45484 secs.
##  1 attributes confirmed important: age_years;
##  42 attributes confirmed unimportant: bacterial_infections_y_1_n_2,
## bmi, comorbidities_1_y_1_n_0, eritrocytes_y_1_n_2, fibrin_y_1_n_2 and
## 37 more;
## Warning in TentativeRoughFix(boruta.lungs_train): There are no Tentative
## attributes! Returning original object.

## Boruta performed 162 iterations in 2.45484 secs.
##  1 attributes confirmed important: age_years;
##  42 attributes confirmed unimportant: bacterial_infections_y_1_n_2,
## bmi, comorbidities_1_y_1_n_0, eritrocytes_y_1_n_2, fibrin_y_1_n_2 and
## 37 more;

# extract the results table in the 'boruta_res' object
extraction <- function(x) {

      ext <- boruta_res[[x]][[1]] %>% 
            
            rownames_to_column(var ='variable') %>% 
           
            select(variable, decision, meanImp) %>% 
            
            filter(decision == 'Confirmed') 
      
    return(ext)
}

extracted_res <- map(1:5, ~extraction(.x))

# create a dataframe of the common boruta results in all the three results lists

boruta_final <- function(extracted_res){

  extraction_merged <- inner_join(extracted_res[[1]], extracted_res[[2]], 
                                    by = 'variable') %>% 
  select(-starts_with('decision'))


  extraction_merged_2 <- inner_join(extraction_merged, extracted_res[[3]], 
                                      by = 'variable') %>% 
  select(-starts_with('decision'))  


  extraction_merged_3 <- inner_join(extraction_merged_2, extracted_res[[4]], 
                                      by = 'variable') %>% 
  select(-starts_with('decision'))  


  extraction_final <- inner_join(extraction_merged_3, extracted_res[[5]], 
                                   by = 'variable') %>% 
  select(-starts_with('decision'))  

  return(extraction_final)

}

# pretty table with the results

boruta_final(extracted_res) |> 
  
  gt() |>  
  
  cols_label(
    variable = md('__Variable__'),
    meanImp.x = md('__MeanImp 1__'),
    meanImp.y = md('__MeanImp 2__'),
    meanImp.x.x = md('__MeanImp 3__'),
    meanImp.y.y = md('__MeanImp 4__'),
    meanImp = md('__MeanImp 5__')
    
  ) |> 
  
    tab_header(
    title = md("**Common important variables**"), #use md() to specify rmkdn formatting
    subtitle = md(glue::glue("_Boruta iteration on 5 seeds_"))
    )
Common important variables
Boruta iteration on 5 seeds
Variable MeanImp 1 MeanImp 2 MeanImp 3 MeanImp 4 MeanImp 5
age_years 4.894423 4.984158 4.952032 4.911679 4.797455

ICU Stay

# creo una funzione per far girare boruta in automatico con tutti i seed
boruta_fun <- function(seed){
  set.seed(seed)
  
  boruta.lungs_train <- Boruta(icu_los~ . , 
         maxRuns = 1000, # aumento il numero per includere le variabili escluse
         data = db_imp |> select(-status_vivo_1_morto_0, -h_los, -ecmo_y_1_n_2))
  print(boruta.lungs_train)
  
  boruta.lungs <- TentativeRoughFix(boruta.lungs_train)
  print(boruta.lungs)
  
  plot <- plot(boruta.lungs, xlab = "", las = 3, cex.axis = 0.8, 
     whichShadow = c(FALSE, FALSE, FALSE),
     pars = par(mar = c(9.5,4,2,1)))

  lungs_df <- attStats(boruta.lungs) %>%
    relocate(decision) %>%
    dplyr::select(-normHits) %>% # così ci sta nella pagina
    arrange(decision, desc(meanImp))

  return(list(lungs_df, plot))
  
  
}

boruta_res <- map(seed, ~ boruta_fun(.x))
## Boruta performed 13 iterations in 0.2984929 secs.
##  No attributes deemed important.
##  43 attributes confirmed unimportant: age_years,
## bacterial_infections_y_1_n_2, bmi, comorbidities_1_y_1_n_0,
## eritrocytes_y_1_n_2 and 38 more;
## Warning in TentativeRoughFix(boruta.lungs_train): There are no Tentative
## attributes! Returning original object.
## Boruta performed 13 iterations in 0.2984929 secs.
##  No attributes deemed important.
##  43 attributes confirmed unimportant: age_years,
## bacterial_infections_y_1_n_2, bmi, comorbidities_1_y_1_n_0,
## eritrocytes_y_1_n_2 and 38 more;
## Boruta performed 13 iterations in 0.2749541 secs.
##  No attributes deemed important.
##  43 attributes confirmed unimportant: age_years,
## bacterial_infections_y_1_n_2, bmi, comorbidities_1_y_1_n_0,
## eritrocytes_y_1_n_2 and 38 more;
## Warning in TentativeRoughFix(boruta.lungs_train): There are no Tentative
## attributes! Returning original object.

## Boruta performed 13 iterations in 0.2749541 secs.
##  No attributes deemed important.
##  43 attributes confirmed unimportant: age_years,
## bacterial_infections_y_1_n_2, bmi, comorbidities_1_y_1_n_0,
## eritrocytes_y_1_n_2 and 38 more;
## Boruta performed 187 iterations in 2.709885 secs.
##  2 attributes confirmed important: comorbidities_1_y_1_n_0,
## il18_delta_ct;
##  41 attributes confirmed unimportant: age_years,
## bacterial_infections_y_1_n_2, bmi, eritrocytes_y_1_n_2, fibrin_y_1_n_2
## and 36 more;
## Warning in TentativeRoughFix(boruta.lungs_train): There are no Tentative
## attributes! Returning original object.

## Boruta performed 187 iterations in 2.709885 secs.
##  2 attributes confirmed important: comorbidities_1_y_1_n_0,
## il18_delta_ct;
##  41 attributes confirmed unimportant: age_years,
## bacterial_infections_y_1_n_2, bmi, eritrocytes_y_1_n_2, fibrin_y_1_n_2
## and 36 more;
## Boruta performed 13 iterations in 0.2905951 secs.
##  No attributes deemed important.
##  43 attributes confirmed unimportant: age_years,
## bacterial_infections_y_1_n_2, bmi, comorbidities_1_y_1_n_0,
## eritrocytes_y_1_n_2 and 38 more;
## Warning in TentativeRoughFix(boruta.lungs_train): There are no Tentative
## attributes! Returning original object.

## Boruta performed 13 iterations in 0.2905951 secs.
##  No attributes deemed important.
##  43 attributes confirmed unimportant: age_years,
## bacterial_infections_y_1_n_2, bmi, comorbidities_1_y_1_n_0,
## eritrocytes_y_1_n_2 and 38 more;
## Boruta performed 13 iterations in 0.2342601 secs.
##  No attributes deemed important.
##  43 attributes confirmed unimportant: age_years,
## bacterial_infections_y_1_n_2, bmi, comorbidities_1_y_1_n_0,
## eritrocytes_y_1_n_2 and 38 more;
## Warning in TentativeRoughFix(boruta.lungs_train): There are no Tentative
## attributes! Returning original object.

## Boruta performed 13 iterations in 0.2342601 secs.
##  No attributes deemed important.
##  43 attributes confirmed unimportant: age_years,
## bacterial_infections_y_1_n_2, bmi, comorbidities_1_y_1_n_0,
## eritrocytes_y_1_n_2 and 38 more;

# extract the results table in the 'boruta_res' object
extraction <- function(x) {

      ext <- boruta_res[[x]][[1]] %>% 
            
            rownames_to_column(var ='variable') %>% 
           
            select(variable, decision, meanImp) %>% 
            
            filter(decision == 'Confirmed') 
      
    return(ext)
}

extracted_res <- map(1:5, ~extraction(.x))

# create a dataframe of the common boruta results in all the three results lists

boruta_final <- function(extracted_res){

  extraction_merged <- inner_join(extracted_res[[1]], extracted_res[[2]], 
                                    by = 'variable') %>% 
  select(-starts_with('decision'))


  extraction_merged_2 <- inner_join(extraction_merged, extracted_res[[3]], 
                                      by = 'variable') %>% 
  select(-starts_with('decision'))  


  extraction_merged_3 <- inner_join(extraction_merged_2, extracted_res[[4]], 
                                      by = 'variable') %>% 
  select(-starts_with('decision'))  


  extraction_final <- inner_join(extraction_merged_3, extracted_res[[5]], 
                                   by = 'variable') %>% 
  select(-starts_with('decision'))  

  return(extraction_final)

}

# pretty table with the results

boruta_final(extracted_res) |> 
  
  gt() |>  
  
  cols_label(
    variable = md('__Variable__'),
    meanImp.x = md('__MeanImp 1__'),
    meanImp.y = md('__MeanImp 2__'),
    meanImp.x.x = md('__MeanImp 3__'),
    meanImp.y.y = md('__MeanImp 4__'),
    meanImp = md('__MeanImp 5__')
    
  ) |> 
  
    tab_header(
    title = md("**Common important variables**"), #use md() to specify rmkdn formatting
    subtitle = md(glue::glue("_Boruta iteration on 5 seeds_"))
    )
Common important variables
Boruta iteration on 5 seeds
Variable MeanImp 1 MeanImp 2 MeanImp 3 MeanImp 4 MeanImp 5

Hospital LOS

# creo una funzione per far girare boruta in automatico con tutti i seed
boruta_fun <- function(seed){
  set.seed(seed)
  
  boruta.lungs_train <- Boruta(h_los~ . , 
         maxRuns = 1000, # aumento il numero per includere le variabili escluse
         data = db_imp |> select(-status_vivo_1_morto_0, -icu_los, -ecmo_y_1_n_2))
  print(boruta.lungs_train)
  
  boruta.lungs <- TentativeRoughFix(boruta.lungs_train)
  print(boruta.lungs)
  
  plot <- plot(boruta.lungs, xlab = "", las = 3, cex.axis = 0.8, 
     whichShadow = c(FALSE, FALSE, FALSE),
     pars = par(mar = c(9.5,4,2,1)))

  lungs_df <- attStats(boruta.lungs) %>%
    relocate(decision) %>%
    dplyr::select(-normHits) %>% # così ci sta nella pagina
    arrange(decision, desc(meanImp))

  return(list(lungs_df, plot))
  
  
}

boruta_res <- map(seed, ~ boruta_fun(.x))
## Warning in addShadowsAndGetImp(decReg, runs): getImp result contains NA(s) or
## NaN(s); replacing with 0(s), yet this is suspicious.
## Boruta performed 355 iterations in 5.093733 secs.
##  2 attributes confirmed important: age_years, il18_delta_ct;
##  41 attributes confirmed unimportant: bacterial_infections_y_1_n_2,
## bmi, comorbidities_1_y_1_n_0, eritrocytes_y_1_n_2, fibrin_y_1_n_2 and
## 36 more;
## Warning in TentativeRoughFix(boruta.lungs_train): There are no Tentative
## attributes! Returning original object.
## Boruta performed 355 iterations in 5.093733 secs.
##  2 attributes confirmed important: age_years, il18_delta_ct;
##  41 attributes confirmed unimportant: bacterial_infections_y_1_n_2,
## bmi, comorbidities_1_y_1_n_0, eritrocytes_y_1_n_2, fibrin_y_1_n_2 and
## 36 more;
## Boruta performed 445 iterations in 10.39789 secs.
##  2 attributes confirmed important: age_years, il18_delta_ct;
##  41 attributes confirmed unimportant: bacterial_infections_y_1_n_2,
## bmi, comorbidities_1_y_1_n_0, eritrocytes_y_1_n_2, fibrin_y_1_n_2 and
## 36 more;
## Warning in TentativeRoughFix(boruta.lungs_train): There are no Tentative
## attributes! Returning original object.

## Boruta performed 445 iterations in 10.39789 secs.
##  2 attributes confirmed important: age_years, il18_delta_ct;
##  41 attributes confirmed unimportant: bacterial_infections_y_1_n_2,
## bmi, comorbidities_1_y_1_n_0, eritrocytes_y_1_n_2, fibrin_y_1_n_2 and
## 36 more;
## Boruta performed 78 iterations in 1.214906 secs.
##  2 attributes confirmed important: age_years, il18_delta_ct;
##  41 attributes confirmed unimportant: bacterial_infections_y_1_n_2,
## bmi, comorbidities_1_y_1_n_0, eritrocytes_y_1_n_2, fibrin_y_1_n_2 and
## 36 more;
## Warning in TentativeRoughFix(boruta.lungs_train): There are no Tentative
## attributes! Returning original object.

## Boruta performed 78 iterations in 1.214906 secs.
##  2 attributes confirmed important: age_years, il18_delta_ct;
##  41 attributes confirmed unimportant: bacterial_infections_y_1_n_2,
## bmi, comorbidities_1_y_1_n_0, eritrocytes_y_1_n_2, fibrin_y_1_n_2 and
## 36 more;
## Boruta performed 320 iterations in 5.964861 secs.
##  2 attributes confirmed important: age_years, il18_delta_ct;
##  41 attributes confirmed unimportant: bacterial_infections_y_1_n_2,
## bmi, comorbidities_1_y_1_n_0, eritrocytes_y_1_n_2, fibrin_y_1_n_2 and
## 36 more;
## Warning in TentativeRoughFix(boruta.lungs_train): There are no Tentative
## attributes! Returning original object.

## Boruta performed 320 iterations in 5.964861 secs.
##  2 attributes confirmed important: age_years, il18_delta_ct;
##  41 attributes confirmed unimportant: bacterial_infections_y_1_n_2,
## bmi, comorbidities_1_y_1_n_0, eritrocytes_y_1_n_2, fibrin_y_1_n_2 and
## 36 more;
## Boruta performed 434 iterations in 7.018322 secs.
##  2 attributes confirmed important: age_years, il18_delta_ct;
##  41 attributes confirmed unimportant: bacterial_infections_y_1_n_2,
## bmi, comorbidities_1_y_1_n_0, eritrocytes_y_1_n_2, fibrin_y_1_n_2 and
## 36 more;
## Warning in TentativeRoughFix(boruta.lungs_train): There are no Tentative
## attributes! Returning original object.

## Boruta performed 434 iterations in 7.018322 secs.
##  2 attributes confirmed important: age_years, il18_delta_ct;
##  41 attributes confirmed unimportant: bacterial_infections_y_1_n_2,
## bmi, comorbidities_1_y_1_n_0, eritrocytes_y_1_n_2, fibrin_y_1_n_2 and
## 36 more;

# extract the results table in the 'boruta_res' object
extraction <- function(x) {

      ext <- boruta_res[[x]][[1]] %>% 
            
            rownames_to_column(var ='variable') %>% 
           
            select(variable, decision, meanImp) %>% 
            
            filter(decision == 'Confirmed') 
      
    return(ext)
}

extracted_res <- map(1:5, ~extraction(.x))

# create a dataframe of the common boruta results in all the three results lists

boruta_final <- function(extracted_res){

  extraction_merged <- inner_join(extracted_res[[1]], extracted_res[[2]], 
                                    by = 'variable') %>% 
  select(-starts_with('decision'))


  extraction_merged_2 <- inner_join(extraction_merged, extracted_res[[3]], 
                                      by = 'variable') %>% 
  select(-starts_with('decision'))  


  extraction_merged_3 <- inner_join(extraction_merged_2, extracted_res[[4]], 
                                      by = 'variable') %>% 
  select(-starts_with('decision'))  


  extraction_final <- inner_join(extraction_merged_3, extracted_res[[5]], 
                                   by = 'variable') %>% 
  select(-starts_with('decision'))  

  return(extraction_final)

}

# pretty table with the results

boruta_final(extracted_res) |> 
  
  gt() |>  
  
  cols_label(
    variable = md('__Variable__'),
    meanImp.x = md('__MeanImp 1__'),
    meanImp.y = md('__MeanImp 2__'),
    meanImp.x.x = md('__MeanImp 3__'),
    meanImp.y.y = md('__MeanImp 4__'),
    meanImp = md('__MeanImp 5__')
    
  ) |> 
  
    tab_header(
    title = md("**Common important variables**"), #use md() to specify rmkdn formatting
    subtitle = md(glue::glue("_Boruta iteration on 5 seeds_"))
    )
Common important variables
Boruta iteration on 5 seeds
Variable MeanImp 1 MeanImp 2 MeanImp 3 MeanImp 4 MeanImp 5
il18_delta_ct 5.208267 5.343406 4.804936 5.214985 5.238157
age_years 3.677935 3.949990 3.975347 4.122035 3.745468

Grafico IL18

Non mi convince molto

cor.test(db_boruta$h_los, db_boruta$il18_delta_ct, method = "spearman") 
## Warning in cor.test.default(db_boruta$h_los, db_boruta$il18_delta_ct, method =
## "spearman"): Cannot compute exact p-value with ties
## 
##  Spearman's rank correlation rho
## 
## data:  db_boruta$h_los and db_boruta$il18_delta_ct
## S = 212.65, p-value = 0.003264
## alternative hypothesis: true rho is not equal to 0
## sample estimates:
##       rho 
## 0.6872749
b <- ggplot(db_boruta, aes(x = h_los, y = il18_delta_ct)) #initiate plot

b + geom_point(aes(color = il18_delta_ct), size = 4) + 
  
  geom_smooth(method = lm) +
  
  scale_color_gradientn(colors = c("#00AFBB", "#E7B800", "#FC4E07")) +
  theme(legend.position = "right") +
  
  theme_classic(base_size = 15) +
  
  xlab(' Hospital Stay (days)') +
  
  ylab('IL-18 expression (ΔCt)') + 
  
  theme(legend.title = element_blank()) +
  
  geom_text(x=90, y=0, label="Rho = 0.68") +
  
  geom_text(aes(fontface = 3), x=90, y=-5, label="p = ") +
  
  geom_text(x=95, y=-5, label="0.003")
## `geom_smooth()` using formula 'y ~ x'

ECMO

# creo una funzione per far girare boruta in automatico con tutti i seed
boruta_fun <- function(seed){
  set.seed(seed)
  
  boruta.lungs_train <- Boruta(ecmo_y_1_n_2~ . , 
         maxRuns = 1000, # aumento il numero per includere le variabili escluse
         data = db_imp |> select(-status_vivo_1_morto_0, -h_los, -icu_los))
  print(boruta.lungs_train)
  
  boruta.lungs <- TentativeRoughFix(boruta.lungs_train)
  print(boruta.lungs)
  
  plot <- plot(boruta.lungs, xlab = "", las = 3, cex.axis = 0.8, 
     whichShadow = c(FALSE, FALSE, FALSE),
     pars = par(mar = c(9.5,4,2,1)))

  lungs_df <- attStats(boruta.lungs) %>%
    relocate(decision) %>%
    dplyr::select(-normHits) %>% # così ci sta nella pagina
    arrange(decision, desc(meanImp))

  return(list(lungs_df, plot))
  
  
}

boruta_res <- map(seed, ~ boruta_fun(.x))
## Warning in addShadowsAndGetImp(decReg, runs): getImp result contains NA(s) or
## NaN(s); replacing with 0(s), yet this is suspicious.
## Boruta performed 999 iterations in 16.29534 secs.
##  2 attributes confirmed important: age_years,
## neutrophilic_granulocytes_percent;
##  40 attributes confirmed unimportant: bacterial_infections_y_1_n_2,
## bmi, comorbidities_1_y_1_n_0, eritrocytes_y_1_n_2, fibrin_y_1_n_2 and
## 35 more;
##  1 tentative attributes left: ifnb1_delta_ct;
## Boruta performed 999 iterations in 16.29534 secs.
## Tentatives roughfixed over the last 999 iterations.
##  2 attributes confirmed important: age_years,
## neutrophilic_granulocytes_percent;
##  41 attributes confirmed unimportant: bacterial_infections_y_1_n_2,
## bmi, comorbidities_1_y_1_n_0, eritrocytes_y_1_n_2, fibrin_y_1_n_2 and
## 36 more;
## Boruta performed 340 iterations in 5.995959 secs.
##  3 attributes confirmed important: age_years,
## neutrophilic_granulocytes_percent, tnf_delta_ct;
##  40 attributes confirmed unimportant: bacterial_infections_y_1_n_2,
## bmi, comorbidities_1_y_1_n_0, eritrocytes_y_1_n_2, fibrin_y_1_n_2 and
## 35 more;
## Warning in TentativeRoughFix(boruta.lungs_train): There are no Tentative
## attributes! Returning original object.

## Boruta performed 340 iterations in 5.995959 secs.
##  3 attributes confirmed important: age_years,
## neutrophilic_granulocytes_percent, tnf_delta_ct;
##  40 attributes confirmed unimportant: bacterial_infections_y_1_n_2,
## bmi, comorbidities_1_y_1_n_0, eritrocytes_y_1_n_2, fibrin_y_1_n_2 and
## 35 more;

## Boruta performed 999 iterations in 15.96097 secs.
##  2 attributes confirmed important: age_years,
## neutrophilic_granulocytes_percent;
##  40 attributes confirmed unimportant: bacterial_infections_y_1_n_2,
## bmi, comorbidities_1_y_1_n_0, eritrocytes_y_1_n_2, fibrin_y_1_n_2 and
## 35 more;
##  1 tentative attributes left: ifnb1_delta_ct;
## Boruta performed 999 iterations in 15.96097 secs.
## Tentatives roughfixed over the last 999 iterations.
##  2 attributes confirmed important: age_years,
## neutrophilic_granulocytes_percent;
##  41 attributes confirmed unimportant: bacterial_infections_y_1_n_2,
## bmi, comorbidities_1_y_1_n_0, eritrocytes_y_1_n_2, fibrin_y_1_n_2 and
## 36 more;
## Boruta performed 246 iterations in 4.002913 secs.
##  4 attributes confirmed important: age_years, ifnb1_delta_ct,
## neutrophilic_granulocytes_percent, tnf_delta_ct;
##  39 attributes confirmed unimportant: bacterial_infections_y_1_n_2,
## bmi, comorbidities_1_y_1_n_0, eritrocytes_y_1_n_2, fibrin_y_1_n_2 and
## 34 more;
## Warning in TentativeRoughFix(boruta.lungs_train): There are no Tentative
## attributes! Returning original object.

## Boruta performed 246 iterations in 4.002913 secs.
##  4 attributes confirmed important: age_years, ifnb1_delta_ct,
## neutrophilic_granulocytes_percent, tnf_delta_ct;
##  39 attributes confirmed unimportant: bacterial_infections_y_1_n_2,
## bmi, comorbidities_1_y_1_n_0, eritrocytes_y_1_n_2, fibrin_y_1_n_2 and
## 34 more;
## Boruta performed 212 iterations in 3.08289 secs.
##  3 attributes confirmed important: age_years,
## neutrophilic_granulocytes_percent, tnf_delta_ct;
##  40 attributes confirmed unimportant: bacterial_infections_y_1_n_2,
## bmi, comorbidities_1_y_1_n_0, eritrocytes_y_1_n_2, fibrin_y_1_n_2 and
## 35 more;
## Warning in TentativeRoughFix(boruta.lungs_train): There are no Tentative
## attributes! Returning original object.

## Boruta performed 212 iterations in 3.08289 secs.
##  3 attributes confirmed important: age_years,
## neutrophilic_granulocytes_percent, tnf_delta_ct;
##  40 attributes confirmed unimportant: bacterial_infections_y_1_n_2,
## bmi, comorbidities_1_y_1_n_0, eritrocytes_y_1_n_2, fibrin_y_1_n_2 and
## 35 more;

# extract the results table in the 'boruta_res' object
extraction <- function(x) {

      ext <- boruta_res[[x]][[1]] %>% 
            
            rownames_to_column(var ='variable') %>% 
           
            select(variable, decision, meanImp) %>% 
            
            filter(decision == 'Confirmed') 
      
    return(ext)
}

extracted_res <- map(1:5, ~extraction(.x))

# create a dataframe of the common boruta results in all the three results lists

boruta_final <- function(extracted_res){

  extraction_merged <- inner_join(extracted_res[[1]], extracted_res[[2]], 
                                    by = 'variable') %>% 
  select(-starts_with('decision'))


  extraction_merged_2 <- inner_join(extraction_merged, extracted_res[[3]], 
                                      by = 'variable') %>% 
  select(-starts_with('decision'))  


  extraction_merged_3 <- inner_join(extraction_merged_2, extracted_res[[4]], 
                                      by = 'variable') %>% 
  select(-starts_with('decision'))  


  extraction_final <- inner_join(extraction_merged_3, extracted_res[[5]], 
                                   by = 'variable') %>% 
  select(-starts_with('decision'))  

  return(extraction_final)

}

# pretty table with the results

boruta_final(extracted_res) |> 
  
  gt() |>  
  
  cols_label(
    variable = md('__Variable__'),
    meanImp.x = md('__MeanImp 1__'),
    meanImp.y = md('__MeanImp 2__'),
    meanImp.x.x = md('__MeanImp 3__'),
    meanImp.y.y = md('__MeanImp 4__'),
    meanImp = md('__MeanImp 5__')
    
  ) |> 
  
    tab_header(
    title = md("**Common important variables**"), #use md() to specify rmkdn formatting
    subtitle = md(glue::glue("_Boruta iteration on 5 seeds_"))
    )
Common important variables
Boruta iteration on 5 seeds
Variable MeanImp 1 MeanImp 2 MeanImp 3 MeanImp 4 MeanImp 5
neutrophilic_granulocytes_percent 6.764996 7.458533 6.814973 7.151645 7.540191
age_years 2.974455 4.404362 3.098043 4.204640 4.409143

Grafico Neutrofili-granulociti

Non mi convince molto neanche questo

ggbetweenstats(
  data = db_boruta,
  x = ecmo_y_1_n_2,
  y = neutrophilic_granulocytes_percent,
  type = 'np'
)

Correlazioni citochine-clinica gruppo COVID

Non ho messo nessun aggiustamento per molteplicità (visto che non lo abbiamo messo neanche prima)

db_cor <- db |> 
  
  filter(group_1_covid_2_transplant_3_donatore == '1') |>
  
  select(age_years, bmi, icu_los, h_los, 
         ifna1_delta_ct, ifna7_delta_ct, ifna8_delta_ct, ifng_delta_ct, il1a_delta_ct, il1b_delta_ct, il9_delta_ct, il18_delta_ct,
         neutrophilic_granulocytes_percent)

ggcorrmat(db_cor,
          type = 'np',
          p.adjust.method = 'none')

Plot significative

Dai plot non sembra niente di allarmante, le correlazioni sembrano più dovute al caso che a qualche effetto biologico

cor.test(db_cor$age_years, db_cor$il18_delta_ct, method = "spearman") 
## Warning in cor.test.default(db_cor$age_years, db_cor$il18_delta_ct, method =
## "spearman"): Cannot compute exact p-value with ties
## 
##  Spearman's rank correlation rho
## 
## data:  db_cor$age_years and db_cor$il18_delta_ct
## S = 1045.1, p-value = 0.03197
## alternative hypothesis: true rho is not equal to 0
## sample estimates:
##        rho 
## -0.5369777
b <- ggplot(db_cor, aes(x = age_years, y = il18_delta_ct)) #initiate plot

b + geom_point(aes(color = il18_delta_ct), size = 4) + 
  
  geom_smooth(method = lm) +
  
  scale_color_gradientn(colors = c("#00AFBB", "#E7B800", "#FC4E07")) +
  theme(legend.position = "right") +
  
  theme_classic(base_size = 15) +
  
  xlab(' Age (years)') +
  
  ylab('IL-18 expression (ΔCt)') + 
  
  theme(legend.title = element_blank()) #+
## `geom_smooth()` using formula 'y ~ x'

  # 
  # geom_text(x=90, y=0, label="Rho = 0.68") +
  # 
  # geom_text(aes(fontface = 3), x=90, y=-5, label="p = ") +
  # 
  # geom_text(x=95, y=-5, label="0.003")


b <- ggplot(db_cor, aes(x = icu_los, y = il18_delta_ct)) #initiate plot

b + geom_point(aes(color = il18_delta_ct), size = 4) + 
  
  geom_smooth(method = lm) +
  
  scale_color_gradientn(colors = c("#00AFBB", "#E7B800", "#FC4E07")) +
  theme(legend.position = "right") +
  
  theme_classic(base_size = 15) +
  
  xlab(' ICU LOS (years)') +
  
  ylab('IL-18 expression (ΔCt)') + 
  
  theme(legend.title = element_blank())
## `geom_smooth()` using formula 'y ~ x'

b <- ggplot(db_cor, aes(x = h_los, y = ifna7_delta_ct)) #initiate plot

b + geom_point(aes(color = ifna7_delta_ct), size = 4) + 
  
  geom_smooth(method = lm) +
  
  scale_color_gradientn(colors = c("#00AFBB", "#E7B800", "#FC4E07")) +
  theme(legend.position = "right") +
  
  theme_classic(base_size = 15) +
  
  xlab(' Hospital LOS (years)') +
  
  ylab('IFN-A7 expression (ΔCt)') + 
  
  theme(legend.title = element_blank())
## `geom_smooth()` using formula 'y ~ x'