Difficulty identifying variable regions in close succession

[eliottpark]

Hi there! Once again, thank you for building this tool, it's hugely useful.

I came across some difficulty when trying to identify variable regions in close succession within the same chain for pretty typical looking VHs and VLs. This is with scfv mode turned on. For one example VH and VL pair, I tried to understand the minimum linker length required to identify both variable regions in the same prediction. Below are the sequences and identified query start and end for the minimum required length of a G4S linker (modulo by increasing length) and the N-1 example.

Is this behavior something that can be handled in the post-processing or is this fully due to model training?

• VH + linker + VL: min linker length = 2

Only found 1 domain for Sequence 1.

VH + G + VL --------------
EVQLVESGGGLVQPGGSLRLSCAASGFTFSSYMMSWVRQAPGKGLEWVATISGTGANTYYPDSVKGRFTISRDNAKNSLYLQMNSLRAEDTAVYYCARQLYYFDYWGQGTTVTVSSGDIQMTQSPSSLSASVGDRVTITCLASQTIGTWLTWYQQKPGKAPKLLIYTATSLADGVPSRFSGSGSGTDFTLTISSLQPEDFATYYCQQVYSIPWTFGGGTKVEIK
Sequence 1-1: start 0, end 115
VH + GG + VL --------------
EVQLVESGGGLVQPGGSLRLSCAASGFTFSSYMMSWVRQAPGKGLEWVATISGTGANTYYPDSVKGRFTISRDNAKNSLYLQMNSLRAEDTAVYYCARQLYYFDYWGQGTTVTVSSGGDIQMTQSPSSLSASVGDRVTITCLASQTIGTWLTWYQQKPGKAPKLLIYTATSLADGVPSRFSGSGSGTDFTLTISSLQPEDFATYYCQQVYSIPWTFGGGTKVEIK
Sequence 1-1: start 0, end 115
Sequence 1-2: start 118, end 224

• VL + linker + VH: min linker length = 6

Only found 1 domain for Sequence 1.

VL + GGGGS + VH --------------
DIQMTQSPSSLSASVGDRVTITCLASQTIGTWLTWYQQKPGKAPKLLIYTATSLADGVPSRFSGSGSGTDFTLTISSLQPEDFATYYCQQVYSIPWTFGGGTKVEIKGGGGSEVQLVESGGGLVQPGGSLRLSCAASGFTFSSYMMSWVRQAPGKGLEWVATISGTGANTYYPDSVKGRFTISRDNAKNSLYLQMNSLRAEDTAVYYCARQLYYFDYWGQGTTVTVSS
Sequence 1-1: start 0, end 106
VL + GGGGSG + VH --------------
DIQMTQSPSSLSASVGDRVTITCLASQTIGTWLTWYQQKPGKAPKLLIYTATSLADGVPSRFSGSGSGTDFTLTISSLQPEDFATYYCQQVYSIPWTFGGGTKVEIKGGGGSGEVQLVESGGGLVQPGGSLRLSCAASGFTFSSYMMSWVRQAPGKGLEWVATISGTGANTYYPDSVKGRFTISRDNAKNSLYLQMNSLRAEDTAVYYCARQLYYFDYWGQGTTVTVSS
Sequence 1-1: start 0, end 106
Sequence 1-2: start 113, end 228

• VH + linker + VH: min linker length = 7

Only found 1 domain for Sequence 1.

VH + GGGGSG + VH --------------
EVQLVESGGGLVQPGGSLRLSCAASGFTFSSYMMSWVRQAPGKGLEWVATISGTGANTYYPDSVKGRFTISRDNAKNSLYLQMNSLRAEDTAVYYCARQLYYFDYWGQGTTVTVSSGGGGSGEVQLVESGGGLVQPGGSLRLSCAASGFTFSSYMMSWVRQAPGKGLEWVATISGTGANTYYPDSVKGRFTISRDNAKNSLYLQMNSLRAEDTAVYYCARQLYYFDYWGQGTTVTVSS
Sequence 1-1: start 0, end 115
VH + GGGGSGG + VH --------------
EVQLVESGGGLVQPGGSLRLSCAASGFTFSSYMMSWVRQAPGKGLEWVATISGTGANTYYPDSVKGRFTISRDNAKNSLYLQMNSLRAEDTAVYYCARQLYYFDYWGQGTTVTVSSGGGGSGGEVQLVESGGGLVQPGGSLRLSCAASGFTFSSYMMSWVRQAPGKGLEWVATISGTGANTYYPDSVKGRFTISRDNAKNSLYLQMNSLRAEDTAVYYCARQLYYFDYWGQGTTVTVSS
Sequence 1-1: start 0, end 115
Sequence 1-2: start 123, end 238

• VL + linker + VL: min linker length = 15

Only found 1 domain for Sequence 1.

VL + GGGGSGGGGSGGGG + VL --------------
DIQMTQSPSSLSASVGDRVTITCLASQTIGTWLTWYQQKPGKAPKLLIYTATSLADGVPSRFSGSGSGTDFTLTISSLQPEDFATYYCQQVYSIPWTFGGGTKVEIKGGGGSGGGGSGGGGDIQMTQSPSSLSASVGDRVTITCLASQTIGTWLTWYQQKPGKAPKLLIYTATSLADGVPSRFSGSGSGTDFTLTISSLQPEDFATYYCQQVYSIPWTFGGGTKVEIK
Sequence 1-1: start 0, end 106
VL + GGGGSGGGGSGGGGS + VL --------------
DIQMTQSPSSLSASVGDRVTITCLASQTIGTWLTWYQQKPGKAPKLLIYTATSLADGVPSRFSGSGSGTDFTLTISSLQPEDFATYYCQQVYSIPWTFGGGTKVEIKGGGGSGGGGSGGGGSDIQMTQSPSSLSASVGDRVTITCLASQTIGTWLTWYQQKPGKAPKLLIYTATSLADGVPSRFSGSGSGTDFTLTISSLQPEDFATYYCQQVYSIPWTFGGGTKVEIK
Sequence 1-1: start 0, end 106
Sequence 1-2: start 122, end 228

[ALGW71]

Hi @eliottpark thanks for your kind words.

Can I just check you are using version 2.0.5 ?

For some reason your first example works for me. Let me know as it is quite possible there is some difference in the way we are running the code?

(test_anarcii) ~/code/ANARCII % anarcii -V           
anarcii 2.0.5

from anarcii import Anarcii

model = Anarcii(seq_type="antibody",
                cpu=False, 
                ncpu=12, 
                mode="accuracy",
                verbose=False)

seq = "EVQLVESGGGLVQPGGSLRLSCAASGFTFSSYMMSWVRQAPGKGLEWVATISGTGANTYYPDSVKGRFTISRDNAKNSLYLQMNSLRAEDTAVYYCARQLYYFDYWGQGTTVTVSSGDIQMTQSPSSLSASVGDRVTITCLASQTIGTWLTWYQQKPGKAPKLLIYTATSLADGVPSRFSGSGSGTDFTLTISSLQPEDFATYYCQQVYSIPWTFGGGTKVEIK"

results = model.number(seq, scfv=True)

print("Length of the sequence:", len(seq))

results = model.number(seq, scfv=True)

for key, value in results.items():
    print(key, ":", value)
    print(seq[value['query_start']: value['query_end']])

Length of the sequence: 224
Sequence-1 : {'numbering': [((1, ' '), 'E'), ((2, ' '), 'V'), ((3, ' '), 'Q'), ((4, ' '), 'L'), ((5, ' '), 'V'), ((6, ' '), 'E'), ((7, ' '), 'S'), ((8, ' '), 'G'), ((9, ' '), 'G'), ((10, ' '), '-'), ((11, ' '), 'G'), ((12, ' '), 'L'), ((13, ' '), 'V'), ((14, ' '), 'Q'), ((15, ' '), 'P'), ((16, ' '), 'G'), ((17, ' '), 'G'), ((18, ' '), 'S'), ((19, ' '), 'L'), ((20, ' '), 'R'), ((21, ' '), 'L'), ((22, ' '), 'S'), ((23, ' '), 'C'), ((24, ' '), 'A'), ((25, ' '), 'A'), ((26, ' '), 'S'), ((27, ' '), 'G'), ((28, ' '), 'F'), ((29, ' '), 'T'), ((30, ' '), 'F'), ((31, ' '), '-'), ((32, ' '), '-'), ((33, ' '), '-'), ((34, ' '), '-'), ((35, ' '), 'S'), ((36, ' '), 'S'), ((37, ' '), 'Y'), ((38, ' '), 'M'), ((39, ' '), 'M'), ((40, ' '), 'S'), ((41, ' '), 'W'), ((42, ' '), 'V'), ((43, ' '), 'R'), ((44, ' '), 'Q'), ((45, ' '), 'A'), ((46, ' '), 'P'), ((47, ' '), 'G'), ((48, ' '), 'K'), ((49, ' '), 'G'), ((50, ' '), 'L'), ((51, ' '), 'E'), ((52, ' '), 'W'), ((53, ' '), 'V'), ((54, ' '), 'A'), ((55, ' '), 'T'), ((56, ' '), 'I'), ((57, ' '), 'S'), ((58, ' '), 'G'), ((59, ' '), 'T'), ((60, ' '), '-'), ((61, ' '), '-'), ((62, ' '), 'G'), ((63, ' '), 'A'), ((64, ' '), 'N'), ((65, ' '), 'T'), ((66, ' '), 'Y'), ((67, ' '), 'Y'), ((68, ' '), 'P'), ((69, ' '), 'D'), ((70, ' '), 'S'), ((71, ' '), 'V'), ((72, ' '), 'K'), ((73, ' '), '-'), ((74, ' '), 'G'), ((75, ' '), 'R'), ((76, ' '), 'F'), ((77, ' '), 'T'), ((78, ' '), 'I'), ((79, ' '), 'S'), ((80, ' '), 'R'), ((81, ' '), 'D'), ((82, ' '), 'N'), ((83, ' '), 'A'), ((84, ' '), 'K'), ((85, ' '), 'N'), ((86, ' '), 'S'), ((87, ' '), 'L'), ((88, ' '), 'Y'), ((89, ' '), 'L'), ((90, ' '), 'Q'), ((91, ' '), 'M'), ((92, ' '), 'N'), ((93, ' '), 'S'), ((94, ' '), 'L'), ((95, ' '), 'R'), ((96, ' '), 'A'), ((97, ' '), 'E'), ((98, ' '), 'D'), ((99, ' '), 'T'), ((100, ' '), 'A'), ((101, ' '), 'V'), ((102, ' '), 'Y'), ((103, ' '), 'Y'), ((104, ' '), 'C'), ((105, ' '), 'A'), ((106, ' '), 'R'), ((107, ' '), 'Q'), ((108, ' '), 'L'), ((109, ' '), 'Y'), ((110, ' '), '-'), ((111, ' '), '-'), ((112, ' '), '-'), ((113, ' '), '-'), ((114, ' '), 'Y'), ((115, ' '), 'F'), ((116, ' '), 'D'), ((117, ' '), 'Y'), ((118, ' '), 'W'), ((119, ' '), 'G'), ((120, ' '), 'Q'), ((121, ' '), 'G'), ((122, ' '), 'T'), ((123, ' '), 'T'), ((124, ' '), 'V'), ((125, ' '), 'T'), ((126, ' '), 'V'), ((127, ' '), 'S'), ((128, ' '), 'S')], 'chain_type': 'H', 'score': 31.011289596557617, 'query_start': 0, 'query_end': 115, 'error': None, 'scheme': 'imgt'}
EVQLVESGGGLVQPGGSLRLSCAASGFTFSSYMMSWVRQAPGKGLEWVATISGTGANTYYPDSVKGRFTISRDNAKNSLYLQMNSLRAEDTAVYYCARQLYYFDYWGQGTTVTVS
Sequence-2 : {'numbering': [((1, ' '), 'D'), ((2, ' '), 'I'), ((3, ' '), 'Q'), ((4, ' '), 'M'), ((5, ' '), 'T'), ((6, ' '), 'Q'), ((7, ' '), 'S'), ((8, ' '), 'P'), ((9, ' '), 'S'), ((10, ' '), 'S'), ((11, ' '), 'L'), ((12, ' '), 'S'), ((13, ' '), 'A'), ((14, ' '), 'S'), ((15, ' '), 'V'), ((16, ' '), 'G'), ((17, ' '), 'D'), ((18, ' '), 'R'), ((19, ' '), 'V'), ((20, ' '), 'T'), ((21, ' '), 'I'), ((22, ' '), 'T'), ((23, ' '), 'C'), ((24, ' '), 'L'), ((25, ' '), 'A'), ((26, ' '), 'S'), ((27, ' '), 'Q'), ((28, ' '), 'T'), ((29, ' '), 'I'), ((30, ' '), '-'), ((31, ' '), '-'), ((32, ' '), '-'), ((33, ' '), '-'), ((34, ' '), '-'), ((35, ' '), '-'), ((36, ' '), 'G'), ((37, ' '), 'T'), ((38, ' '), 'W'), ((39, ' '), 'L'), ((40, ' '), 'T'), ((41, ' '), 'W'), ((42, ' '), 'Y'), ((43, ' '), 'Q'), ((44, ' '), 'Q'), ((45, ' '), 'K'), ((46, ' '), 'P'), ((47, ' '), 'G'), ((48, ' '), 'K'), ((49, ' '), 'A'), ((50, ' '), 'P'), ((51, ' '), 'K'), ((52, ' '), 'L'), ((53, ' '), 'L'), ((54, ' '), 'I'), ((55, ' '), 'Y'), ((56, ' '), 'T'), ((57, ' '), 'A'), ((58, ' '), '-'), ((59, ' '), '-'), ((60, ' '), '-'), ((61, ' '), '-'), ((62, ' '), '-'), ((63, ' '), '-'), ((64, ' '), '-'), ((65, ' '), 'T'), ((66, ' '), 'S'), ((67, ' '), 'L'), ((68, ' '), 'A'), ((69, ' '), 'D'), ((70, ' '), 'G'), ((71, ' '), 'V'), ((72, ' '), 'P'), ((73, ' '), '-'), ((74, ' '), 'S'), ((75, ' '), 'R'), ((76, ' '), 'F'), ((77, ' '), 'S'), ((78, ' '), 'G'), ((79, ' '), 'S'), ((80, ' '), 'G'), ((81, ' '), '-'), ((82, ' '), '-'), ((83, ' '), 'S'), ((84, ' '), 'G'), ((85, ' '), 'T'), ((86, ' '), 'D'), ((87, ' '), 'F'), ((88, ' '), 'T'), ((89, ' '), 'L'), ((90, ' '), 'T'), ((91, ' '), 'I'), ((92, ' '), 'S'), ((93, ' '), 'S'), ((94, ' '), 'L'), ((95, ' '), 'Q'), ((96, ' '), 'P'), ((97, ' '), 'E'), ((98, ' '), 'D'), ((99, ' '), 'F'), ((100, ' '), 'A'), ((101, ' '), 'T'), ((102, ' '), 'Y'), ((103, ' '), 'Y'), ((104, ' '), 'C'), ((105, ' '), 'Q'), ((106, ' '), 'Q'), ((107, ' '), 'V'), ((108, ' '), 'Y'), ((109, ' '), 'S'), ((110, ' '), '-'), ((111, ' '), '-'), ((112, ' '), '-'), ((113, ' '), '-'), ((114, ' '), 'I'), ((115, ' '), 'P'), ((116, ' '), 'W'), ((117, ' '), 'T'), ((118, ' '), 'F'), ((119, ' '), 'G'), ((120, ' '), 'G'), ((121, ' '), 'G'), ((122, ' '), 'T'), ((123, ' '), 'K'), ((124, ' '), 'V'), ((125, ' '), 'E'), ((126, ' '), 'I'), ((127, ' '), 'K'), ((128, ' '), '-')], 'chain_type': 'K', 'score': 30.634769439697266, 'query_start': 117, 'query_end': 223, 'error': None, 'scheme': 'imgt'}
DIQMTQSPSSLSASVGDRVTITCLASQTIGTWLTWYQQKPGKAPKLLIYTATSLADGVPSRFSGSGSGTDFTLTISSLQPEDFATYYCQQVYSIPWTFGGGTKVEI

[eliottpark]

Hi @ALGW71 , thank you for the quick reply. I am using 2.0.5, and was using the speed mode previously which seems to be the source of discrepancy. With accuracy mode on, it seems that the VH + VL, VL + VL, and VL + VH combinations work with no linker between variable regions. However, I am still seeing that VH + VH required an 8 residue long linker in order for the model to capture the second VH, at least for this sequence set. Would you be able to check if you see that on your end as well?
Thanks.

from anarcii import Anarcii

model = Anarcii(seq_type="antibody",
                cpu=False, 
                ncpu=12, 
                mode="accuracy",
                verbose=False)

vh = 'EVQLVESGGGLVQPGGSLRLSCAASGFTFSSYMMSWVRQAPGKGLEWVATISGTGANTYYPDSVKGRFTISRDNAKNSLYLQMNSLRAEDTAVYYCARQLYYFDYWGQGTTVTVSS'
vl = 'DIQMTQSPSSLSASVGDRVTITCLASQTIGTWLTWYQQKPGKAPKLLIYTATSLADGVPSRFSGSGSGTDFTLTISSLQPEDFATYYCQQVYSIPWTFGGGTKVEIK'

count = 0
for i in range(0, 11):
     linker = ''.join(['GGGGS'[k%5] for k in range(0, i)])
     seqs = [vh + linker + vh]
     out = model.number(seqs, scfv=True)
     if len(out.keys()) > 1:
         count += 1
         if count > 1:
             break
     print(f'VH + {linker} + VH --------------')  
     print(seqs[0])
     for key, dict in out.items():
         print(f'{key}: start {dict["query_start"]}, end {dict["query_end"]}')

[ALGW71]

I am still seeing that VH + VH required an 8 residue long linker in order for the model to capture the second VH, at least for this sequence set. Would you be able to check if you see that on your end as well?

This is still an issue. Reopening.