From what I understand, you are looking to filter the elements of each column that meet a certain condition and then present pairs of columns and values that comply.

The data

  

It is not good practice to show the example data as an image, since we can not directly load that data to R to validate a response. Here are the readable data in R (I passed an OCR, may have errors)

library(tidyverse)
tribble(
~FV, ~HC, ~HT, ~JA, ~JN, ~JP, ~JU, ~LA, ~LC, ~LM, ~MG,
0, 0, 0.03, 0, 0, 0, 0, 0, 0, 0, 0.1,
0, 0, 0.02, 0, 0, 0, 0, 0, 0, 0, 0.11,
0, 0, 0.04, 0, 0, 0, 0, 0, 0, 0, 0.11,
0, 0, 0.03, 0, 0, 0, 0, 0, 0, 0, 0.11,
0, 0, 0.02, 0, 0, 0, 0, 0, 0, 0, 0.12,
0, 0, 0.03, 0, 0, 0, 0, 0, 0, 0, 0.1,
0, 0, 0.03, 0, 0, 0, 0, 0, 0, 0, 0.11) -> datos_de_user

Procedure

To obtain the result you are looking for, we are going to do three steps, using the library functions tidyverse .

Solution

#library(tidyverse)           # Está más arriba, pero es necesario.

datos_de_user %>% 
  gather(clave, valor) %>%    # Paso a formato largo clave-valor
  group_by(clave) %>%         # Agrupo por claves.
  filter(valor > 0.01)        # Filtro los valores mayores a 0.1

# A tibble: 14 x 2
# Groups:   clave [2]
clave valor
<chr> <dbl>
1  HT     0.03
2  HT     0.02
3  HT     0.04
4  HT     0.03
5  HT     0.02
6  HT     0.03
7  HT     0.03
8  MG     0.1 
9  MG     0.11
10 MG     0.11
11 MG     0.11
12 MG     0.12
13 MG     0.1 
14 MG     0.11

Another example with better column names.

datos_de_user %>% 
  gather(variable, probabilidad) %>% 
  group_by(variable) %>% 
  filter(probabilidad > 0.01 & probabilidad < 0.1) #Uso & para más de una condición

# A tibble: 7 x 2
# Groups:   variable [1]
variable probabilidad
<chr>           <dbl>
1 HT               0.03
2 HT               0.02
3 HT               0.04
4 HT               0.03
5 HT               0.02
6 HT               0.03
7 HT               0.03

Update after clarification

I understand that what interests you is knowing which orders have columns that meet a certain condition.

In that case you could do it with a logic similar to the previous case, just that you should:

  

Note2: these are the results with the test data you shared with us. With other data or other conditions you should have other results, perhaps only one of the orders meets the condition. In any case, the logic is the same.

resultados <- datos_de_user %>%                        #Para asignarle un nombre al resultado. 
      mutate(orden = paste("PB", 4184:4190, sep = "")) %>% #Agrego unos números de orden
      gather(clave, valor, -orden) %>%                     #-orden para "bloquear" esa variable.
      group_by(orden, clave) %>%                           #doble agrupamiento, para conservar orden
      filter(valor > 0.01)  %>%                            #filtro por condición
      arrange(orden) %>%                                   #orden para que cada orden quede junta 
      select(-valor)                                       #saco la columna "valor"

resultados

# A tibble: 14 x 2
# Groups:   orden, clave [14]
orden  clave
<chr>  <chr>
1 PB4184 HT   
2 PB4184 MG   
3 PB4185 HT   
4 PB4185 MG   
5 PB4186 HT   
6 PB4186 MG   
7 PB4187 HT   
8 PB4187 MG   
9 PB4188 HT   
10 PB4188 MG   
11 PB4189 HT   
12 PB4189 MG   
13 PB4190 HT   
14 PB4190 MG 

Undoubtedly the solution of @mpaladino using dplyr is more elegant and simple, however something similar with base R can also be achieved. The idea is basically the same:

• We convert the "wide" format to a "long" one
• Then we simply filter the desired values

Solution:

library(tidyverse)

tribble(
    ~Orden, ~FV, ~HC, ~HT, ~JA, ~JN, ~JP, ~JU, ~LA, ~LC, ~LM, ~MG,
    'PB4184', 0, 0, 0.03, 0, 0, 0, 0, 0, 0, 0, 0.1,
    'PB4185', 0, 0, 0.02, 0, 0, 0, 0, 0, 0, 0, 0.11,
    'PB4186', 0, 0, 0.04, 0, 0, 0, 0, 0, 0, 0, 0.11,
    'PB4187', 0, 0, 0.03, 0, 0, 0, 0, 0, 0, 0, 0.11,
    'PB4188', 0, 0, 0.02, 0, 0, 0, 0, 0, 0, 0, 0.12,
    'PB4189', 0, 0, 0.03, 0, 0, 0, 0, 0, 0, 0, 0.1,
    'PB4190', 0, 0, 0.03, 0, 0, 0, 0, 0, 0, 0, 0.11) -> datos_de_user


newdata <- reshape(data=datos_de_user, 
                   direction="long",
                   varying=list(2:12),
                   times=names(datos_de_user)[2:12],
                   timevar="Columna",
                   v.names="probabilidad"
)
newdata <- newdata[newdata$probabilidad > 0.02 & newdata$probabilidad <= 0.1,]    
newdata <- newdata[order(newdata$Orden),c(1,2,3)]
newdata

• With reshape() we modify the original "wide" format to a "long" one
• With newdata[newdata$probabilidad > 0.02 & newdata$probabilidad <= 0.1,] we filter the probabilities according to some desired criteria
• Finally and as a further detail, with newdata[order(newdata$Orden),c(1,2,3)] we sort everything by the original row number and also rearrange the columns

The only peculiarity of this solution is that we keep the Orden corresponding to each row:

# A tibble: 7 x 3
  Orden  Columna probabilidad
  <chr>  <chr>          <dbl>
1 PB4184 HT              0.03
2 PB4184 MG              0.1 
3 PB4186 HT              0.04
4 PB4187 HT              0.03
5 PB4189 HT              0.03
6 PB4189 MG              0.1 
7 PB4190 HT              0.03

There are some rows / orders that are repeated since there is more than one column in which the condition we are evaluating is given. The consistent way to handle these cases is to duplicate these rows as we do in the previous example, however, for your comments you want each row to be unique and the matches to be displayed in a "horizontal" way, it is not a convenient way to leave the data , but it can be practical at the time of visualizing them, to do this we can execute the following:

final <- aggregate( paste(Columna, probabilidad, sep='=') ~ Orden, newdata, c)
names(final)[2] <- 'probabilidad'
final

   Orden    probabilidad
1 PB4184 HT=0.03, MG=0.1
2 PB4186         HT=0.04
3 PB4187         HT=0.03
4 PB4189 HT=0.03, MG=0.1
5 PB4190         HT=0.03

Here we see that we are left with only one row per Orden and a new column probabilidad with the cases that have coincided with the previous criterion. In this case the column probabilidad is a vector of chains of variable size.