1) Why don't you grep 'LTR' lines from both the files

grep -w "LTR" FileA.bed > FileA_greppedLTR.bed

grep -w "LTR" FileB.bed > FileB_greppedLTR.bed

2) Use bed tools on grepped files

using sqlite3

$ awk 'BEGIN {printf("create table if not exists TA (C TEXT,S int,E int, T text); BEGIN TRANSACTION;\n");} { printf("insert into TA(C,S,E,T) values (\"%s\",%s,%s,\"%s\");\n",$1,$2,$3,$4);} END {printf("commit;\n");}' fileA | sqlite3 db.sqlite
$ awk 'BEGIN {printf("create table if not exists TB (C TEXT,S int,E int, T text); BEGIN TRANSACTION;\n");} { printf("insert into TB(C,S,E,T) values (\"%s\",%s,%s,\"%s\");\n",$1,$2,$3,$4);} END {printf("commit;\n");}' fileB | sqlite3 db.sqlite
$ sqlite3 db.sqlite 'select * from TA,TB where TA.C=TB.C and TA.T=TB.T and NOT(TA.E <= TB.S or TB.E <= TA.S)'

Sorry I could not able to add as a reply to your comment because of the format of the message. I am writing it in the answer section

Example input (BED tab-delimited):

scaffold_1 1026586 1027586 Mariner

scaffold_1 925211 935660 DNA_transposon

scaffold_1 925880 926300 LTR

scaffold_1 11111 22324 DNA_transposon

scaffold_1 1026586 1027586 Mariner

scaffold_1 21432 21444 Mariner

SCRIPT ( splitFile.sh ) :

file=$1

cols=$(cat $file | cut -f4 | awk '!x[$0]++' | tr '\n' ' ' | sed -e 's/(.*) /\1/g');

for name in $cols

{

IFS='/' read -a file_name <<< "$file";

fname=$(echo $file_name | cut -f 1 -d '.')

grep -w "$name" $file > ${fname}_${name}.bed

}

Run script:

./splitFile.sh FileB.bed

Output files:

1) FileB_DNA_transposon.bed

scaffold_1 925211 935660 DNA_transposon

scaffold_1 11111 22324 DNA_transposon

2) FileB_LTR.bed

scaffold_1 925880 926300 LTR

3) FileB_Mariner.bed

scaffold_1 1026586 1027586 Mariner

scaffold_1 1026586 1027586 Mariner

scaffold_1 21432 21444 Mariner

Run this script for FileA.bed and then compare appropriate pairs using BedTools. This is just a suggestion (idea), you can modify according to your needs.

Generating intermediate files is unnecessary. Doing pairwise comparisons is unnecessary.

The bedops toolkit supports multiple standard input streams, so for one "condition" or ID field, you can use a process substitution to generate a stream of subsetted elements, wherever you would have otherwise specified a path to a regular file ("filename"):

$ bedops --whatever-set-operation <(awk '$4 == "DNA_transposon"' A.bed) <(awk '$4 == "DNA_transposon"' B.bed) > answer.DNA_transposon.bed

More generally:

$ id="DNA_transposon" && bedops --whatever-set-operation <(awk -vid=$id '$4==id' A.bed) <(awk -vid=$id '$4==id' B.bed) > answer.$id.bed

From here, it is easy to deal with automating the even more general case of multiple IDs.

Just set up a loop that walks through a list of unique IDs from one of the inputs via cut, sort anduniq:

$ for id in `cut -f4 A.bed | sort | uniq`; do bedops --whatever-set-operation <(awk -vid=$id '$4==id' A.bed) <(awk -vid=$id '$4==id' B.bed) > answer.$id.bed; done

If you want a list of IDs from two or more files, just use cat to concatenate them to build the ID list:

$ for id in `cat A.bed B.bed | cut -f4 | sort | uniq`; do bedops --whatever-set-operation <(awk -vid=$id '$4==id' A.bed) <(awk -vid=$id '$4==id' B.bed) > answer.$id.bed; done

Thanks to default support for Unix input streams in bedops tools and to functionality in the bash shell, all of these approaches generate no intermediate (regular) files and should run pretty quickly.