the target-chain flags always take the chain from the first whatever you input

[Qmi3]

I noticed that if i want to feed the templates for other chains in target sequence, they are always aligned to first k sequence, is it possible to fix this bug

[clami66]

Hi @Qmi3 ,

Is this a protein containing multiple identical sequences? In that case you need to use the flag --separate_homomer_msas when running run_alphafold.py (see example in the README) so that alphafold is forced to read a separate template for each identical chain instead of defaulting to the first.

We will make this flag the default in our next release, but right now you need to explicitly use it.

Let me know if this fixes the issue, otherwise it might be good to know a bit more about your inputs and commands for further debugging

[Qmi3]

thanks for your reply, i want to give the templates for the fourth and fifth sequence in the target fasta, but i used this command,
python prepare_templates.py --target examples/7yqh/smc56_nse256.fasta
--template examples/7yqh/7yqh_AB.pdb
--output_dir AF_models/
--align
--target_chains D E
--template_chains A B
it maps the template A B chain to the first two sequence in the target fasta, the target fasta contains five chain, this is the align output:
DO ALIGN blast
KRVKIAKP--DLSSFQPGSII-KIRLQDFVTYTLTEFNLS-PSLNMIIGPNGSGKSTFVCAVCLGLAGKPEYIGRSKKVEDFIK-NGQDVSKIEITLKNSPNVTDIEYIDARDETIKITRIITRSKRRSDYLIND-YQVSESVVKTLVAQLNIQLDNLCQFLS-QERVEEFARLKSV-KLLVE-TIRSIDA-SLLD-VLDELRELQGN-EQSLQK--DLD---FKKAKIVHLRQES--DKLRKSVESLRD-FQNKKGEIELHSQLLPYVKVKDHKEKLNIYKEEYERAKANLRAILKDKKPFANTKKTLENQVEELTEKCSLKTDEFLKAK--EKINEIFEKLNTIR-DEVIKKKNQNEYYR-GRTKKLQATIIST-KEDFLRSQEI-LAQ---THLPEK-SVFEDIDIKRKEIINKEGEIRD-LISEIDAKANAINHEMRSIQRQAESKTKSLTTTDKIGILNQDQDLKEVRDAVLMVREHPE-MK--DKILEPPIMTVSAINAQFAAYLAQCVDYNTSKALTVV-DSDSYKLFA-NPIL-DKFKVNLRELSSADTTPPVPAETVRDLGF--EGYLSDFITGDKRVMKML-CQ--TS-KIHTIPVSRR-ELT-PAQIKKLITPRP--NGKILFKRIIHGNRLVDIKQSAYGSKQVFPTDVSIKQTNFYQGSIMSNEQKIRIENEIINLKNEYNDRKSTLDALSNQKSGYRHELSELASKNDDINREAHQLNEIRKKYTMRKSTIETLREKLD--QLKREARKDVSQKIKDIDDQIQQLLLKQRHLLSKMASSMKSLKNCQKELISTQILQFEAQNMDVSMNDVIGFFNEREADLKSQYEDKKKFVKEMRDTPEFQSW-MREIRSYDQDT-KEKLNKVAEKYEEEGNFNLSFVQDVLDKLESEIAMVNHDESAVTILDQVTAELRELEHTVPQQSKDLETIKAKLKEDHAVLEPKLDDIVSKISARFARLFNNVGSAGAVRLEKPKDYAEWKIEIMVKFRDNAPLKKLDSHTQSGGERAVSTVLYMIALQEFTSAPFRVVDEINQGMDSRNERIVHKAMVENACAENTSQYFLITPKLLTGLHYHEKMRIHCVMAGSWIPNPSEDPKMIHFGETSNYSFD
--M--A--LND-N---P--I-PK----------------SVP-LH----P----KS--------G---K--Y---------F--HN----------L----------H--ARD--L--------S--------N-IYQ-----------Q--------C-Y--KQ--IDE-----TINQL-VDST--S---PS---TI--------G-IEE--Q-VADI-TSTYK---L--L---STYE----S-ES--NSF-----D-E-H------IK--D----L---K------K-N----------F---K-----Q----S---S---D----A-CPQ-I-D----LST--WD---K------Y-RTG--E-L--T--A-PK---L-S-ELYL--NMPT--PE-PA---------------------TMV-------N--N----T-----D--T--L----KI--L------K-V----L-----P-YI-WND----P---T--------------CV---------I-PD-----L--QNP--AD-------E----D-----------DL--QIEG-------G-K-I-E-LTC-PIT-CK----P----YE--AP-----LIS-R-KCN---------H----V------------F--D---R--D---G-I----Q-----N--------Y------L-----Q--GY----------------------------T----T----R---DCPQ----A-----------------------------A--------C-----S-QV---------VSM---------R--D----------FV---RD-P-----IM-ELR------CK-----IA-K--------M--------K-ES----------------Q---E---------Q---D----K-----------------------R---------S---------------------------------S--Q-----A-------I-------------D-------------V-----------------L-----------------------------------------------
DO ALIGN blast
IEQVDDE--LLSLTAQQENEEQQQQRKRRRHQFAPMTLEESPSGYIKKVIL--RNFMC--HEHF-EL-E--LGSRLN-F--IVGNNGSGKSAI-LTAITIGLGAKASETNRGSSLK-D--LIR--EGC-YSA-KIIL-HLDNSKYG-AYQQG-I-FGNEII-VERIIKRDGPASF--S-LRSENG--KE-I--SNKKKDIQTVVDYFSVPVS-NPMCFL-SQDAARSFL-TA-STSQDKYSHFMKGTLLQEITENLLY-ASAIH----DSAQENMALHLE-NLKSLKAEYEDA-KKLLR-E--LNQTS-D-LNE--RKMLLQAKSLWIDVAHNTDACKNLENEISGIQQKV-DEVTEKIRN-RQEKIERYT-SD-G-TTIEAQIDAKVIYVNEKDSEHQNAR----ELLRDV--KSRFEKEKSN--QAEAQSN-I---DQGRKKVDALNKTIAHLEEELTKEMGGDKDQMRQELEQLE--KA-NEKLR--E-VNNSLVV-SLQDVK-NEERD-IQHERESELRT-I--SRSI-QNKKV-----E-LQNIAKGNDTFLMNFD--RNM-DRLLRT--I--E--QRKNEFETPAIGPLGSLVTIRKG--FEKWTRSIQRAIS-SS-LNA-FVV--S-NP-KD--NRLFR-D-I-MRSCGIRSNIPIVTYC-L-SQFD-YSKGRA---H----G-NYPTIVDALE-FSK-PE-IECLFVDLSRIERIV-LIED-KNEARNFLQRNPVNVNMALSLRDRRS--G--FQLSGGYRLDTVTYQDKIRLKVNSSS-DNGTQYLKDLIEQET-KELQNIR-DRYEEKLSEVRSRLKEID-G--R-LK-STK-NE-M--RKTNFR-MTELKMNVGKVVDTGILNSKINERKNQEQAIASYEAAKEELGLK-IEQIAQEAQPIKEQYDSTKLALVEAQDE--LQQLK-EDINSRQSKIQKYKDDTIYY-EDKKK-VY--LEN-IKKIEVNVAALK-EGIQRQIQNACAFCSKERIENVDLPDTQEE-IKRE-LDKVSRMIQKAEKSLGLSQEEVI--ALF-EKCRNKYKEGQKK--YMEIDEALNRLHNSLKARDQNYKNAEKGTCFDAD--MDFRASLKV--RKFS-----GNLSF--IK-DT-KS--LE-IY-ILTTNDEKARNVDTLSGGEKSFS-QMALLLATW-KPMRSRIIALDEFDVFMDQVNRKIGTTLIVKKLK-D--IARTQTI--IITPQDI-GKIA-DID-SSGVSIHRMRDP
M---D--GALI-------N--------------S-V-L------Y---V--SPRN---GAH-YFVELTEKHL---L-AFEML--N--S----MCL------L-----E-N------YDHVLL-FLE-CQF--GK---SH--N----LA----VIPF--DIILV--L--------FTLSTL-SE--YYKEPILRAN---D----------P--YN-----TS----R---ET-LS--R-R-A--LK--LLQK------YLA--I-LKEFDS--E------QYNL------Y-D-LE-LLRCQFFL---AIDTL--TPKK---Q-K--W------------------G-----FD------R-FR-----R-TKSESGVT-----------Y---R----QNA-SVDPEL--D-QAKT-F-K---NPYR----S-YISCLEQ-R------N-TI--L--------G-N----R--L--L-NLK-LNE---PGEFIN--MILWTL----SN----SLQ-E--S---TPLFLS-S-HE---IWMPLLEIL--I----D--L--F-SCR--QD-----YFIQHEVAQ--N------V----S-----K-SLF------VQR-LSES-PL-AVF--FESLN-TRNFANR-F-SEYVFL-NC----D-----Y-KLPS--DNY----ATPVHPVYNGEN--TIVD---TY--IP-TIKC-----S------PL---YKS------QKS-----LAL--R-RK-LIGSCFKL-----L--------LR--V----PD-G--H-R-LI---TPR-I--V-AD--D-----V------I-QGISRTL-AS--FNDILQFKK--F-FMTE---N---------L-S-------QE----SY----------FI--------P---------L-L--A--EGTL----SE-I------L-K--D-T---QE----CV-VIL--TL--VE-N---L-SDGV-----S---FC------N-------E-VI---GL--V-----KS-K----------CFA-FTEQC-S-----Q--ASY-E--EAV--L-N-I----------EK--C-D--VCL-----L-VLLR-Y-LLHLIG--T-EAI-LD-AK-EQLEMLHAI-----EK--N-D--SG--R---RQ-------WAK-------AL----------N--LG---------NDPPL-----LYPIVS-Q-MFG-V-HD--KS--V-I--I-E-
Run AlphaFold with, e.g.:
python run_alphafold.py --fasta_paths examples/7yqh/smc56_nse256.fasta --flagfile databases.flag --flagfile AF_models/smc_test/smc56_nse256/template_data/templates.flag --output_dir AF_models/smc_test --cross_chain_templates --dropout --model_preset='multimer_v2' --separate_homomer_msas

[clami66]

Hi again, I think I understand what you mean now. The quickest and easiest way to fix this is, I think, to put the 4th and 5th sequence to the top of the fasta file so that they are 1st and 2nd. Then call with --target_chains A B . You might also have to move the MSAs around so that they are in the correct folders

I have been looking into more explicit support for partial templates, it should come when I am done putting together a jupyter/colab notebook for the pipeline.

Anyway, let me know if this works or if you run into more issues, as this is a useful test case and it should definitely be fixed!

[clami66]

Hi @Qmi3 I have included explicit support for partial templates now, see instructions here

In your example, where the template covers chaind D and E it should be enough to run, e.g:

python prepare_templates.py --target examples/7yqh/smc56_nse256.fasta
--template examples/7yqh/7yqh_AB.pdb
--output_dir AF_models/
--target_chains A B C D E
--template_chains - - - A B

Notice that you should not use the --align flag in this case.

I hope you have the possibility to try this, let me know if it works in that case.

[Qmi3]

Thanks for your kindly reply, when i make partial templates,i get 5 pdb_seqres_*.txt under the folder of AF_models/9a06/template_data/, but when i run prediction, it ran into error:
Traceback (most recent call last):
File "run_alphafold.py", line 512, in
app.run(main)
File "/lustre/grp/gyqlab/zhangcw/miniconda3/envs/pytorch_1/lib/python3.8/site-packages/absl/app.py", line 308, in run
_run_main(main, args)
File "/lustre/grp/gyqlab/zhangcw/miniconda3/envs/pytorch_1/lib/python3.8/site-packages/absl/app.py", line 254, in _run_main
sys.exit(main(argv))
File "run_alphafold.py", line 385, in main
template_searcher = hmmsearch.Hmmsearch(
File "/lustre/grp/gyqlab/zhangcw/AF_unmasked/alphafold/data/tools/hmmsearch.py", line 63, in init
if not os.path.exists(self.database_path):
File "/lustre/grp/gyqlab/zhangcw/miniconda3/envs/pytorch_1/lib/python3.8/genericpath.py", line 19, in exists
os.stat(path)
TypeError: stat: path should be string, bytes, os.PathLike or integer, not list

AF_models/9a06/template_data/templates.flag shows it has 5 path for the tempalte_database, is there any method to fix this error?

[Qmi3]

This is what template.flag look like:
--cross_chain_templates
--template_mmcif_dir=/lustre/grp/gyqlab/zhangcw/AF_unmasked/AF_models/9a06/template_data/mmcif_files
--pdb_seqres_database_path=/lustre/grp/gyqlab/zhangcw/AF_unmasked/AF_models/9a06/template_data/pdb_seqres_A.txt,
/lustre/grp/gyqlab/zhangcw/AF_unmasked/AF_models/9a06/template_data/pdb_seqres_B.txt,/lustre/grp/gyqlab/zhangcw/AF_unmasked/AF_models/9a06/template_data/pdb_seqres_C.txt,/lustre/grp/gyqlab/zhangcw/AF_unmasked/AF_models/9a06/template_data/pdb_seqres_D.txt,/lustre/grp/gyqlab/zhangcw/AF_unmasked/AF_models/9a06/template_data/pdb_seqres_E.txt
--use_precomputed_msas

[clami66]

That's how they should look. Have you updated the whole repository? The alphafold code has been modified to allow for separate pdb_seqres_* files