So I've got a solution, though it's rather convoluted and likely not the most efficient!

Using a shorter example of different food type data:

> df
   Type       Food
1 Fruit      Apple
2 Fruit       Pear
3 Fruit     Banana
4 Fruit     Orange
5   Veg     Carrot
6   Veg     Potato
7   Veg    Parsnip
8 Pulse   Chickpea
9 Pulse Broad Bean

First, use pivot_wider (from tidyR). The line below incorporates an answer from here. It's necessary to create a unique identifier for each Type, otherwise the output contains a list-column.

> df_pivot <- df %>% group_by(Type) %>% mutate(row = row_number()) %>% tidyr::pivot_wider(names_from = Type, values_from = Food) %>% select(-row)

This should produce a tibble with NAs where the gaps are.

> df_pivot
# A tibble: 4 × 3
  Fruit  Veg     Pulse     
  <chr>  <chr>   <chr>     
1 Apple  Carrot  Chickpea  
2 Pear   Potato  Broad Bean
3 Banana Parsnip NA        
4 Orange NA      NA

Next, this needs to be transposed so the columns become rows again:

> transposed_df_pivot <-as.data.frame(t(as.matrix(df_pivot)))
> transposed_df_pivot
            V1         V2      V3     V4
Fruit    Apple       Pear  Banana Orange
Veg     Carrot     Potato Parsnip   <NA>
Pulse Chickpea Broad Bean    <NA>   <NA>

For those interested in finalising the formatting ready for TopGO, I ended up then exporting transposed_df_pivot as a csv, removed the column names in excel, then did the rest in python, as below:

>>>Import pandas a pd

# load data with no headers
>>> df = pd.read_csv('your-exported-R-dataframe.csv', header=None)

# Create a new column 'combined', containing all values from columns 1-4 in each row, separated by commas, while also ignoring NaNs
>>> df["combined"] = df[df.columns[1:62]].apply(lambda x: ', '.join(x[x.notnull()]), axis = 1)

# Combine the columns you need into a new df and export as a tab-delimited file with .map file extension
>>> df2 = df[[0, 'combined']]
>>> df2.to_csv('Your_file_name.map', sep = '\t', index=False, header=False)