How Can I Sort Blast Output Based On E-Value?

Question

[biostar]

Hello: I have a blast output with about 50000 sequence list. The problem I have is that I need to sort those results based on the smallest e-value to greater e-value. Could you guys please suggest me anything that I can resort to? Thanks! The blast output looks like this:

    Tx_332323_c0_seq1    110815944    RecName: Full=Gag-Pol polyprotein; AltName: Full=Pr160Gag-Pol; Contains: RecName: Full=Matrix protein p17; Short=MA; Contains: RecName: Full=Capsid protein p24; Short=CA; Contains: RecName: Full=Spacer peptide p2; Contains: RecName: Full=Nucleocapsid protein p7; Short=NC; Contains: RecName: Full=Transframe peptide; Short=TF; Contains: RecName: Full=p6-pol; Short=p6*; Contains: RecName: Full=Protease; AltName: Full=PR; AltName: Full=Retropepsin; Contains: RecName: Full=Reverse transcriptase/ribonuclease H; AltName: Full=Exoribonuclease H; AltName: Full=p66 RT; Contains: RecName: Full=p51 RT; Contains: RecName: Full=p15; Contains: RecName: Full=Integrase; Short=IN    1446    5    457    107    1028    1138    4e-06    66    35.00    -3/0
Tv_333332_c0_seq2    206558251    RecName: Full=Zinc finger MYM-type protein 5    656    8    41    796    405    615    6e-14    92    27.63    2/0
Tv_144391_c0_seq1    75057844    RecName: Full=Nitrogen permease regulator 2-like protein; Short=NPR2-like protein; AltName: Full=Tumor suppressor candidate 4    380    4    1410    304    7    380    6e-118    65    48.27    -3/0
Ty_116400_c0_seq1    73920872    RecName: Full=Longitudinals lacking protein, isoforms F/I/K/T    970    1    29    190    902    956    4e-09    33    45.45    2/0
Ty_144400_c0_seq2    73920872    RecName: Full=Longitudinals lacking protein, isoforms F/I/K/T    970    1    29    190    902    956    2e-08    20    45.45    2/0
Tx_444402_c0_seq1    74834619    RecName: Full=Cathepsin L-like proteinase; Flags: Precursor    324    10    57    1034    1    323    9e-75    91    42.09    3/0

[Pierre Lindenbaum]

you know, there is a linux tool named 'sort': It sorts.

[biostar]

Thanks, but I tried that and gives faulty result.

[Andrzej Zielezinski]

Try this way: sort -k 10,10g input -o output.

[PoGibas]

Also, there is always the manual way...

Answer 1

[PoGibas]

As Pierre suggested you can try sort (take a look at -g option (for general numerical value)).

With a given example this works:

awk '{print $(NF-3),$0}' INPUT | sort -gk1  
# Used awk as original INPUT needs a little bit munging

[biostar]

Thanks Pgibas! The awk command works as needed!

Answer 2

[Michael]

Don't sort by e-value, sort by score instead! The reason is that there can be many more ties in e-values due to numerical accuracy (e-value is exponentially decreasing wrt score) and database sizes. Often, one will for example get many alignments with e-value of 0 but different scores. Based on the numerical value of e-value these will be ranked randomly. Except for ties yielding random ranking, sorting by bit-score of HSPs of the same query and subject, will yield the same ranking of HSPs as sorting by e-values.

[biostar]

Thanks Mike, good to know that as well.