Difference between Pyrodigal 2.0.4 and Prodigal 2.6.3

[pchaumeil]

Hello,

I have run another big test to compare prodigal and pyrodigal across ~400K genomes.

from this test ~17K genomes have a difference in their gene calling between the 2 softwares.
I have attached the list of all genomes with a difference here :
index_genomes.txt

here are few examples of these differences:

for GCA_934838455.1:

diff prodigal_2.6.3_amino.faa pyrodigal_2.0.4_amino.faa
>CAKWEX010000332.1_3 # 2936 # 3754 # -1 # ID=332_3;partial=00;start_type=ATG;rbs_motif=AGGA;rbs_spacer=5-10bp;gc_cont=0.602
MSNHFEGLGKTWLTLLNDPEKEVPAVVMQVMKEGKTRDCWQRKDSKEETMVLAWPVETGF
RAGVTVHGNAGDQLRPVSTYPLLEGAPNDMTVNETYLWQNETEGEVSATCNEGANPLWFY
SPFLFRDRENLTPGVRHTFLIAGLAYGLRRALLDEMTITEGVEYERYVAEWLAQNPGKTR
LDVPQLTVDLRGARIVVPGDVASEYQIRVPVTSVEEMHIQNEKIYMLIVEFGLNTPNPLR
FPLYAPERVCKIVPQAGDEIDAIIWLQGRIID*
---
>CAKWEX010000332.1_3 # 2936 # 3763 # -1 # ID=332_3;partial=01;start_type=Edge;rbs_motif=None;rbs_spacer=None;gc_cont=0.603
IGPMSNHFEGLGKTWLTLLNDPEKEVPAVVMQVMKEGKTRDCWQRKDSKEETMVLAWPVE
TGFRAGVTVHGNAGDQLRPVSTYPLLEGAPNDMTVNETYLWQNETEGEVSATCNEGANPL
WFYSPFLFRDRENLTPGVRHTFLIAGLAYGLRRALLDEMTITEGVEYERYVAEWLAQNPG
KTRLDVPQLTVDLRGARIVVPGDVASEYQIRVPVTSVEEMHIQNEKIYMLIVEFGLNTPN
PLRFPLYAPERVCKIVPQAGDEIDAIIWLQGRIID*

for GCA_934561095.1

diff prodigal_2.6.3_amino.faa pyrodigal_2.0.4_amino.faa
>CAKTGG010000246.1_2 # 738 # 2552 # -1 # ID=246_2;partial=00;start_type=TTG;rbs_motif=AGGAGG;rbs_spacer=5-10bp;gc_cont=0.618
MKLSARKDVPVNETWDLSLIFAAEADFEAAVEKTKALADTLEKTYKNALTTPESIAECLA
LYEELEILLYQTTSYTSLAVSVDYTDTEAQKKDAKMSALAAEIGSRLSFIESEIADAPEE
LIRAAMDKTERAKHYLAEILREKPHRLSAETEKVLAALRPVFNAPYDIYHMTKLADMKFD
SFTVNSKEYPLGYSLFEDEYEYEADTDVRRAAFRAFSDKLRQYENTTAATYNTYLTQQRI
MAHQRGFADMFEADLFADHVTREMYDRQIDLITEKLAPAMRKYARLVGKMNKLDRVTFAD
LKLPLDAEFDPRVTIEESREYVRSALSVLGQDYADMVDEAYDKRWIDFARNAGKETGGFC
SSPYGCNSFILLSWNNRMADVFTIAHELGHAGHFRLCNGAQSLFDTNVSGYLIEAPSTMN
ELLLAQDLLRKNTDKRFRRWVLSSLIGHTYYHNFVTHLREAWYQREAMNIIEQGGAVNAE
TLSGIFRRNLETFWGDAVELTEGCELTWMRQPHYYMGLYSYTYSAGLTLATQAALNIAAE
GESAVARWRAMLEAGSTRGPLGLAEIAGIDLSTPDALEHTIAYISDIIDEIAVLTEELDG
ITLD*
---
>CAKTGG010000246.1_2 # 738 # 2564 # -1 # ID=246_2;partial=01;start_type=Edge;rbs_motif=None;rbs_spacer=None;gc_cont=0.616
EDIHLKLSARKDVPVNETWDLSLIFAAEADFEAAVEKTKALADTLEKTYKNALTTPESIA
ECLALYEELEILLYQTTSYTSLAVSVDYTDTEAQKKDAKMSALAAEIGSRLSFIESEIAD
APEELIRAAMDKTERAKHYLAEILREKPHRLSAETEKVLAALRPVFNAPYDIYHMTKLAD
MKFDSFTVNSKEYPLGYSLFEDEYEYEADTDVRRAAFRAFSDKLRQYENTTAATYNTYLT
QQRIMAHQRGFADMFEADLFADHVTREMYDRQIDLITEKLAPAMRKYARLVGKMNKLDRV
TFADLKLPLDAEFDPRVTIEESREYVRSALSVLGQDYADMVDEAYDKRWIDFARNAGKET
GGFCSSPYGCNSFILLSWNNRMADVFTIAHELGHAGHFRLCNGAQSLFDTNVSGYLIEAP
STMNELLLAQDLLRKNTDKRFRRWVLSSLIGHTYYHNFVTHLREAWYQREAMNIIEQGGA
VNAETLSGIFRRNLETFWGDAVELTEGCELTWMRQPHYYMGLYSYTYSAGLTLATQAALN
IAAEGESAVARWRAMLEAGSTRGPLGLAEIAGIDLSTPDALEHTIAYISDIIDEIAVLTE
ELDGITLD*

Is this the normal to have that many difference across all these genomes? Is Pyrodigal more accurate in this case?

Thank you

[althonos]

Thanks @pchaumeil, I'll have a look!

Is this the normal to have that many difference across all these genomes? Is Pyrodigal more accurate in this case?

I can't say for sure before finding out what's causing the discrepancy, the bug may come either from Prodigal or from Pyrodigal. From the examples you provided it looks like the issue is coming from RBS detection on the reverse strand edge, which I though I fixed as a cause of #27.

[althonos]

I think I found a bug, and it's actually coming from Prodigal:

In your first gene, there is a start codon located around 10bp of the contig edge on the reverse strand. Prodigal / Pyrodigal will look for the RBS motif, and will scan the region before the ATG codon: GGATAGGCCCCATG. The very beginning of the contig is GGA, which is a potential RBS.

Now, here's the difference:

• Pyrodigal will reach the contig edge, and figure the motif to be RBS motif to be GGA. This is a low-scoring motif (2.35) as is only partially matches an RBS consensus sequence. Therefore, the start codon at index 12 gets a low RBS score, and is discarded.
• Prodigal will reach the contig edge, but continue reading past the edge, causing an out-of-buffer access. This may cause it to think there is a A after the contig end, and to mark the RBS motif as being AGGA. Because this is a high-scoring motif (7.85), the start codon is retained as the start of the gene.

As far as I can tell the bug only occurs on the reverse strand, which may explain why your two examples (and perhaps the rest of your discrepancies) all occur on a partial gene on the reverse strand.

[althonos]

Indeed, if you reverse-complement the contig and try again, both Pyrodigal and Prodigal agree on the following gene:

>CAKWEX010000332.1_1 # 3 # 830 # 1 # ID=1_1;partial=10;start_type=Edge;rbs_motif=None;rbs_spacer=None;gc_cont=0.601
IGPMSNHFEGLGKTWLTLLNDPEKEVPAVVMQVMKEGKTRDCWQRKDSKEETMVLAWPVE
TGFRAGVTVHGNAGDQLRPVSTYPLLEGAPNDMTVNETYLWQNETEGEVSATCNEGANPL
WFYSPFLFRDRENLTPGVRHTFLIAGLAYGLRRALLDEMTITEGVEYERYVAEWLAQNPG
KTRLDVPQLTVDLRGARIVVPGDVASEYQIRVPVTSVEEMHIQNEKIYMLIVEFGLNTPN
PLRFPLYAPERVCKIVPQAGDEIDAIIWLQGRIID*

which has an Edge start, so this means Prodigal accurately recognized the RBS motif as GGA this time, and discarded the start codon as expected. I'll make a bug report on the original repository 👍

[althonos]

I'm gonna close this, given I think this was linked to hyattpd/Prodigal#100. Feel free to open another issue if you keep saying this problem with the new version 👍