Screening insertion locations with PoET

In this tutorial, we’ll use PoET to evaluate the acceptibility of insertion locations for the nuclear localization signal PAAKRVKLD from c-Myc into a target protein.

%matplotlib inline

import numpy as np
import matplotlib
import matplotlib.pyplot as plt
import json

import openprotein
import openprotein.fasta as fasta

Connect to the OpenProtein.AI web server

with open('secrets.config', 'r') as f:
    config = json.load(f)

session = openprotein.connect(config['username'], config['password'])

Define the peptide to insert and the protein to analyze

We’ll examine insertion of the nuclear localization signal peptide from c-Myc.

nls = b'PAAKRVKLD'

Examine insertion sites in aliphatic amidase

name = 'AMIE_PSEAE'
wt = b'MRHGDISSSNDTVGVAVVNYKMPRLHTAAEVLDNARKIAEMIVGMKQGLPGMDLVVFPEYSLQGIMYDPAEMMETAVAIPGEETEIFSRACRKANVWGVFSLTGERHEEHPRKAPYNTLVLIDNNGEIVQKYRKIIPWCPIEGWYPGGQTYVSEGPKGMKISLIICDDGNYPEIWRDCAMKGAELIVRCQGYMYPAKDQQVMMAKAMAWANNCYVAVANAAGFDGVYSYFGHSAIIGFDGRTLGECGEEEMGIQYAQLSLSQIRDARANDQSQNHLFKILHRGYSGLQASGDGDRGLAECPFEFYRTWVTDAEKARENVERLTRSTTGVAQCPVGRLPYEGLEKEA'

print('>' + name)
for i in range(0, len(wt), 80):
    print(wt[i:i+80].decode())

>AMIE_PSEAE
MRHGDISSSNDTVGVAVVNYKMPRLHTAAEVLDNARKIAEMIVGMKQGLPGMDLVVFPEYSLQGIMYDPAEMMETAVAIP
GEETEIFSRACRKANVWGVFSLTGERHEEHPRKAPYNTLVLIDNNGEIVQKYRKIIPWCPIEGWYPGGQTYVSEGPKGMK
ISLIICDDGNYPEIWRDCAMKGAELIVRCQGYMYPAKDQQVMMAKAMAWANNCYVAVANAAGFDGVYSYFGHSAIIGFDG
RTLGECGEEEMGIQYAQLSLSQIRDARANDQSQNHLFKILHRGYSGLQASGDGDRGLAECPFEFYRTWVTDAEKARENVE
RLTRSTTGVAQCPVGRLPYEGLEKEA

Create the MSA and define an ensemble prompt

To run PoET, we use the OpenProtein API to construct an MSA from homologues of the wildtype sequence using homology search and then define an ensemble prompt containing 10 prompts sampled with default parameters. Query sequences will then be scored conditioned on the fitness landscape represented by the sequences in the prompt.

OpenProtein uses an asynchronous API, where potentially long running functions return a job ID that can be used to query for completed results. The wait_until_done function can be used to poll for completion.

# search for homologs to automatically create an MSA for the seed sequence
msa = session.align.create_msa(wt)
print(msa)
msa.wait_until_done(verbose=True)

# create the prompt, set the seed for reproducibility
prompt = msa.sample_prompt(num_ensemble_prompts=10, random_seed=1)
print(prompt)
prompt.wait_until_done(verbose=True)

status=<JobStatus.SUCCESS: 'SUCCESS'> job_id='ca5a70d8-7072-4ac7-be88-ed7fee9f04be' job_type=<JobType.align_align: '/align/align'> created_date=datetime.datetime(2024, 4, 1, 7, 56, 11, 651228) start_date=None end_date=datetime.datetime(2024, 4, 1, 7, 56, 11, 651688) prerequisite_job_id=None progress_message=None progress_counter=None num_records=None sequence_length=None msa_id='ca5a70d8-7072-4ac7-be88-ed7fee9f04be'

Waiting: 100%|██████████| 100/100 [00:00<00:00, 7615.21it/s, status=SUCCESS]

status=<JobStatus.PENDING: 'PENDING'> job_id='a2b5509c-51a2-4fea-9daf-aeaa3f2ecad2' job_type=<JobType.align_prompt: '/align/prompt'> created_date=datetime.datetime(2024, 4, 1, 7, 56, 11, 687433) start_date=None end_date=None prerequisite_job_id=None progress_message=None progress_counter=None num_records=None sequence_length=None msa_id='a2b5509c-51a2-4fea-9daf-aeaa3f2ecad2' prompt_id='a2b5509c-51a2-4fea-9daf-aeaa3f2ecad2'

Waiting: 100%|██████████| 100/100 [00:21<00:00,  4.55it/s, status=SUCCESS]

True

Enumerate and score the insertions

poet = session.embedding.get_model("poet")

queries = [wt]
for i in range(len(wt)):
    q = wt[:i] + nls + wt[i+1:]
    queries.append(q)
len(queries)

347

future_scores = poet.score(prompt=prompt, sequences=queries)
print(future_scores.job)
future_scores.wait_until_done(verbose=True)

status=<JobStatus.SUCCESS: 'SUCCESS'> job_id='62ab8201-3446-493b-8b43-0f685d130fe2' job_type=<JobType.poet_score: '/poet'> created_date=datetime.datetime(2024, 3, 25, 7, 13, 37, 216000) start_date=datetime.datetime(2024, 3, 25, 7, 20, 5, 490139) end_date=datetime.datetime(2024, 3, 25, 7, 20, 22, 155155) prerequisite_job_id=None progress_message=None progress_counter=None num_records=None sequence_length=None parent_id=None s3prefix=None page_size=None page_offset=None num_rows=None result=None n_completed=None

Waiting: 100%|██████████| 100/100 [00:00<00:00, 5304.55it/s, status=SUCCESS]

True

results = future_scores.get()
scores_ensemble = np.stack([r[2] for r in results], axis=0)
scores_ensemble.shape

(347, 10)

baseline = scores_ensemble[0]
scores = scores_ensemble[1:].mean(axis=-1) - baseline.mean()

_, ax = plt.subplots(figsize=(20, 4))
ax.scatter(np.arange(1, len(wt)+1), scores)
ax.set_xlim(0, len(wt)+2)
plt.xlabel('Position')
plt.ylabel('PoET Likelihood Score')
plt.title('Insertion Site Tolerance Scan')

Text(0.5, 1.0, 'Insertion Site Tolerance Scan')

# sort out the most favorable insertion sites
order = np.argsort(-scores)
print('Score   ', 'Position', sep='\t')
for i in order[:10]:
    print(f'{scores[i]:8.5f}', i+1, sep='\t')

Score           Position
-28.64021       346
-29.65063       345
-30.90579       344
-31.50744       342
-31.81487       343
-33.61958       341
-34.40652       339
-34.64433       338
-35.18435       340
-36.36819       2

Sensibly, PoET predicts that the C-terminus is likely to be the most tolerated site to insert the NLS peptide. The N-terminus may also be tolerated.

Examine insertion sites in Cas9

from io import BytesIO

fasta_string = b'''>sp|J7RUA5|CAS9_STAAU CRISPR-associated endonuclease Cas9 OS=Staphylococcus aureus OX=1280 GN=cas9 PE=1 SV=1
MKRNYILGLDIGITSVGYGIIDYETRDVIDAGVRLFKEANVENNEGRRSKRGARRLKRRR
RHRIQRVKKLLFDYNLLTDHSELSGINPYEARVKGLSQKLSEEEFSAALLHLAKRRGVHN
VNEVEEDTGNELSTKEQISRNSKALEEKYVAELQLERLKKDGEVRGSINRFKTSDYVKEA
KQLLKVQKAYHQLDQSFIDTYIDLLETRRTYYEGPGEGSPFGWKDIKEWYEMLMGHCTYF
PEELRSVKYAYNADLYNALNDLNNLVITRDENEKLEYYEKFQIIENVFKQKKKPTLKQIA
KEILVNEEDIKGYRVTSTGKPEFTNLKVYHDIKDITARKEIIENAELLDQIAKILTIYQS
SEDIQEELTNLNSELTQEEIEQISNLKGYTGTHNLSLKAINLILDELWHTNDNQIAIFNR
LKLVPKKVDLSQQKEIPTTLVDDFILSPVVKRSFIQSIKVINAIIKKYGLPNDIIIELAR
EKNSKDAQKMINEMQKRNRQTNERIEEIIRTTGKENAKYLIEKIKLHDMQEGKCLYSLEA
IPLEDLLNNPFNYEVDHIIPRSVSFDNSFNNKVLVKQEENSKKGNRTPFQYLSSSDSKIS
YETFKKHILNLAKGKGRISKTKKEYLLEERDINRFSVQKDFINRNLVDTRYATRGLMNLL
RSYFRVNNLDVKVKSINGGFTSFLRRKWKFKKERNKGYKHHAEDALIIANADFIFKEWKK
LDKAKKVMENQMFEEKQAESMPEIETEQEYKEIFITPHQIKHIKDFKDYKYSHRVDKKPN
RELINDTLYSTRKDDKGNTLIVNNLNGLYDKDNDKLKKLINKSPEKLLMYHHDPQTYQKL
KLIMEQYGDEKNPLYKYYEETGNYLTKYSKKDNGPVIKKIKYYGNKLNAHLDITDDYPNS
RNKVVKLSLKPYRFDVYLDNGVYKFVTVKNLDVIKKENYYEVNSKCYEEAKKLKKISNQA
EFIASFYNNDLIKINGELYRVIGVNNDLLNRIEVNMIDITYREYLENMNDKRPPRIIKTI
ASKTQSIKKYSTDILGNLYEVKSKKHPQIIKKG'''

with BytesIO(fasta_string) as f:
    names, sequences = fasta.parse(f)
name = names[0].decode()
wt = sequences[0]

print('>' + name)
for i in range(0, len(wt), 80):
    print(wt[i:i+80].decode())

>sp|J7RUA5|CAS9_STAAU CRISPR-associated endonuclease Cas9 OS=Staphylococcus aureus OX=1280 GN=cas9 PE=1 SV=1
MKRNYILGLDIGITSVGYGIIDYETRDVIDAGVRLFKEANVENNEGRRSKRGARRLKRRRRHRIQRVKKLLFDYNLLTDH
SELSGINPYEARVKGLSQKLSEEEFSAALLHLAKRRGVHNVNEVEEDTGNELSTKEQISRNSKALEEKYVAELQLERLKK
DGEVRGSINRFKTSDYVKEAKQLLKVQKAYHQLDQSFIDTYIDLLETRRTYYEGPGEGSPFGWKDIKEWYEMLMGHCTYF
PEELRSVKYAYNADLYNALNDLNNLVITRDENEKLEYYEKFQIIENVFKQKKKPTLKQIAKEILVNEEDIKGYRVTSTGK
PEFTNLKVYHDIKDITARKEIIENAELLDQIAKILTIYQSSEDIQEELTNLNSELTQEEIEQISNLKGYTGTHNLSLKAI
NLILDELWHTNDNQIAIFNRLKLVPKKVDLSQQKEIPTTLVDDFILSPVVKRSFIQSIKVINAIIKKYGLPNDIIIELAR
EKNSKDAQKMINEMQKRNRQTNERIEEIIRTTGKENAKYLIEKIKLHDMQEGKCLYSLEAIPLEDLLNNPFNYEVDHIIP
RSVSFDNSFNNKVLVKQEENSKKGNRTPFQYLSSSDSKISYETFKKHILNLAKGKGRISKTKKEYLLEERDINRFSVQKD
FINRNLVDTRYATRGLMNLLRSYFRVNNLDVKVKSINGGFTSFLRRKWKFKKERNKGYKHHAEDALIIANADFIFKEWKK
LDKAKKVMENQMFEEKQAESMPEIETEQEYKEIFITPHQIKHIKDFKDYKYSHRVDKKPNRELINDTLYSTRKDDKGNTL
IVNNLNGLYDKDNDKLKKLINKSPEKLLMYHHDPQTYQKLKLIMEQYGDEKNPLYKYYEETGNYLTKYSKKDNGPVIKKI
KYYGNKLNAHLDITDDYPNSRNKVVKLSLKPYRFDVYLDNGVYKFVTVKNLDVIKKENYYEVNSKCYEEAKKLKKISNQA
EFIASFYNNDLIKINGELYRVIGVNNDLLNRIEVNMIDITYREYLENMNDKRPPRIIKTIASKTQSIKKYSTDILGNLYE
VKSKKHPQIIKKG

Create the MSA and define an ensemble prompt

# search for homologs to automatically create an MSA for the seed sequence
msa = session.align.create_msa(wt)
print(msa)
msa.wait_until_done(verbose=True)

# create the prompt, set the seed for reproducibility
prompt = msa.sample_prompt(num_ensemble_prompts=10, random_seed=1)
print(prompt)
prompt.wait_until_done(verbose=True)

status=<JobStatus.SUCCESS: 'SUCCESS'> job_id='100703c8-93e2-438d-980b-5da3a654f240' job_type=<JobType.align: '/align/align'> created_date=datetime.datetime(2024, 3, 25, 7, 20, 31, 541591) start_date=None end_date=datetime.datetime(2024, 3, 25, 7, 20, 31, 541835) prerequisite_job_id=None progress_message=None progress_counter=None num_records=None sequence_length=None msa_id='100703c8-93e2-438d-980b-5da3a654f240'

Waiting: 100%|██████████| 100/100 [00:00<00:00, 7171.34it/s, status=SUCCESS]

status=<JobStatus.PENDING: 'PENDING'> job_id='2f87c58c-0ca2-4e66-9c6b-6b145ac7adbe' job_type=<JobType.align_prompt: '/align/prompt'> created_date=datetime.datetime(2024, 3, 25, 7, 20, 31, 580251) start_date=None end_date=None prerequisite_job_id=None progress_message=None progress_counter=None num_records=None sequence_length=None msa_id='2f87c58c-0ca2-4e66-9c6b-6b145ac7adbe' prompt_id='2f87c58c-0ca2-4e66-9c6b-6b145ac7adbe'

Waiting: 100%|██████████| 100/100 [02:31<00:00,  1.52s/it, status=SUCCESS]

True

Enumerate and score the insertions

queries = [wt]
for i in range(len(wt)):
    q = wt[:i] + nls + wt[i+1:]
    queries.append(q)
len(queries)

1054

future_scores = poet.score(prompt=prompt, sequences=queries)
print(future_scores.job)
future_scores.wait_until_done(verbose=True)

status=<JobStatus.SUCCESS: 'SUCCESS'> job_id='3dbefbed-d7f3-431c-a4d8-ed9e1cbcf182' job_type=<JobType.poet_score: '/poet'> created_date=datetime.datetime(2024, 3, 25, 7, 23, 8, 8314) start_date=datetime.datetime(2024, 3, 25, 7, 28, 24, 513588) end_date=datetime.datetime(2024, 3, 25, 7, 29, 15, 154194) prerequisite_job_id=None progress_message=None progress_counter=None num_records=None sequence_length=None parent_id=None s3prefix=None page_size=None page_offset=None num_rows=None result=None n_completed=None

Waiting: 100%|██████████| 100/100 [00:00<00:00, 6865.44it/s, status=SUCCESS]

True

results = future_scores.get()
scores_ensemble = np.stack([r[2] for r in results], axis=0)
scores_ensemble.shape

(1054, 10)

baseline = scores_ensemble[0]
scores = scores_ensemble[1:].mean(axis=-1) - baseline.mean()

_, ax = plt.subplots(figsize=(20, 4))
ax.scatter(np.arange(1, len(wt)+1), scores)
ax.set_xlim(0, len(wt)+2)
plt.xlabel('Position')
plt.ylabel('PoET Likelihood Score')
plt.title('Insertion Site Tolerance Scan')

Text(0.5, 1.0, 'Insertion Site Tolerance Scan')

# sort out the most favorable insertion sites
order = np.argsort(-scores)
print('Score   ', 'Position', sep='\t')
for i in order[:10]:
    print(f'{scores[i]:8.5f}', i+1, sep='\t')

Score           Position
-28.56653       732
-29.34781       1053
-29.71613       738
-29.94740       733
-30.11392       742
-30.16401       731
-30.36705       737
-30.62352       740
-31.28050       736
-31.69111       739

For Cas9, PoET predicts that the best tolerated insertion site is at position 732 or position 738. The C-terminus is also predited to be amongst the best tolerated. Examining the plot also shows that the N-terminus should also be tolerated. Engineered Cas9 proteins have used both N- and C-terminal nuclear localization signals, but the internal 732/738 sites are novel predictions.

Examining the crystal structure of this Cas9, 5AXW, shows that the the 732-738 stretch is a surface exposed loop, a plausible insertion location.