Dry Lab

At NTU Dry Lab, we envisioned a project to alter T7 RNA Polymerase (RNAP) sequences from the Wild Type, and configure alternate forms of the protein.

Project Overview

We recognized early on that our goal was adventurous and bold, and far too large to accomplish in one massive program. We knew that to have any hopes of success, we had to break the project down to its simpler parts. So, we structured our approach in three key components:

1. Build and Train a Model

  • Ancestral sequence data was used as the training set for the model we built.

    • This model was rigorously evaluated for its performance using HuberLoss, achieving a minimised loss of 0.1542. While this value is formidable for an alpha program, it falls short of our target of 0.1000. To enhance our assessment, we adapted our code to generate alternate RNAP sequences, enabling comparative analyses against the ancestral data.

  • The model demonstrated significant promise, particularly with additional time and an expanded datapool. To ensure our project was robust, the team opted not to rely on an alpha version of our model. So we turned to ProGen2.

    • 2. Generate Alternate RNAP Sequences

  • ProGen2 is a far more robust and superior model. Harnessing the powers of ProGen2 through HuggingFace, the second component of the project came to life: generation of alternative RNAP sequences became possible.
  • The 5 alternate RNA sequences produced are as follows:
    •     1. MNTINIAKNDFSDIELAAIPFNTLADHYGERLAREQLALEHESYEMGEARFRKMFERQLKAGEVADNAAAKPLITTLLPKMIA
    SGKTTWFEEVKAKRGKRPTAFQFLQEIKPEAVAYITIKTTLACLTSADNTTVQAVASAIGRAIEDEARFGRIRDLEAKHFKKNV
    EEQLRLIKEHVYKKAFMQVVEADMLSKGLLGGEAWSSWHKEDSIHVGVRCIEMLIESTGMVSLHRQNAGVVGQDSETIELA
    PEYAEAIATRAESLLDISPMFQPCVVPPKPWTGITGGGYWANGRRPLALVRTHSKKALMRYEDVYMPEVYKAINIAQNTAW
    KINKKVLAVANVITKWKHSSFKAIPAIEREELPMKPEDIDMNPEALTAWKRAAAAVYRKDKARKSRRISLEFMLEQANKFANH
    KAIWFPYNMDWRGRVYAVSMFNPDAPKTTKGLLTLAKGKPIGKEGYYWLKIHGANCAGVDKVPFPERIKFIEENHENIMAC
    AKSPLENTWWAEQDSPFCFLAFCFEYAGVQSFVKSYNCSLPLAFDGSCSGIQHFSAMLRDEVGGRAVNLLPSETVQDIYGI
    VAKKVNEILQADAINGTDNEVVTVTDENTGEISEKVKKLLVKLAGQWLAYGVTRSVTKRSVMTLAYGSKEFGFRQQVLEDTI
    QPAIDSGKGLMFTQPNQAAGYMAKLIWESVSVTVVAAVEAMNWARRRGKLLAAEVKDKKTGEILRKRCAVHWVTPDGFPV
    WQEYKKPIQTRLNLMFLGQFRLQPTINTNKDSEIDAHKQESGIAPNFVHSQIIEKSRKTVVWAHEKYGIESFALIHDSFGTIPA
    DAANLFKAVRETMVDTYESCDVLADFYDQFADQLHESQLDKMPALPAKGNLNLRDILESDSEAVE

    2. MNTINIAKNDFSDIELAAIPFNTLADHYGERLAREQLALEHESYEMGEARFRKMFERQLKAGEVADNAAAKPLITTLLPKMIE
    AGKTPWFEEVKAKRGKRPTAFQFLQEIKPEAVAYITIKTTLACLTSADNTTVQAVASAIGRAIEDEARFGRIRDLEAKHFKKNV
    EEQLAKLEKHVYKKAFMQVVEADMLSKGLLGGEAWSSWHKEDSIHVGVRCIEMLIESTGMVSLHRQNAGVVGQDSETIEL
    APEYAEAIATRAADVLAISPMFQPCVVPPKPWTGITGGGYWANGRRPLALVRTHSKKALMRYEDVYMPEVYKAINIAQNTA
    WKINKKVLAVANVITKWKHKEGLSIPAIEREELPMKPEDIDMNPEALTAWKRAAAAVYRKDKARKSRRISLEFMLEQANKFAN
    HKAIWFPYNMDWRGRVYAVSMFNPSELKETKGLLTLAKGKPIGKEGYYWLKIHGANCAGVDKVPFPERIKFIEENHENIMAC
    AKSPLENTWWAEQDSPFCFLAFCFEYAGVQLVKKGYNCSLPLAFDGSCSGIQHFSAMLRDEVGGRAVNLLPSETVQDIYGI
    VAKKVNEILQADAINGTDNEVVTVTDENTGEISEKVKEVAVDLAGQWLAYGVTRSVTKRSVMTLAYGSKEFGFRQQVLEDTI
    QPAIDSGKGLMFTQPNQAAGYMAKLIWESVSVTVVAAVEAMNWRTAAAKLLAAEVKDKKTGEILRKRCAVHWVTPDGFPV
    WQEYKKPIQTRLNLMFLGQFRLQPTINTNKDSEIDAHKQESGIAPNFVHSQSSSGSRKTVVWAHEKYGIESFALIHDSFGTIP
    ADAANLFKAVRETMVDTYESCDVLADFYDQFADQLHESQLDKMPALPAKGNLNLRDILESDSGDLY

    3. MNTINIAKNDFSDIELAAIPFNTLADHYGERLAREQLALEHESYEMGEARFRKMFERQLKAGEVADNAAAKPLITTLLPKMIR
    RLEDGWFEEVKAKRGKRPTAFQFLQEIKPEAVAYITIKTTLACLTSADNTTVQAVASAIGRAIEDEARFGRIRDLEAKHFKKNV
    EEQLKKRLKHVYKKAFMQVVEADMLSKGLLGGEAWSSWHKEDSIHVGVRCIEMLIESTGMVSLHRQNAGVVGQDSETIEL
    APEYAEAIATRAASLVRISPMFQPCVVPPKPWTGITGGGYWANGRRPLALVRTHSKKALMRYEDVYMPEVYKAINIAQNTA
    WKINKKVLAVANVITKWKHGEKKTIPAIEREELPMKPEDIDMNPEALTAWKRAAAAVYRKDKARKSRRISLEFMLEQANKFAN
    HKAIWFPYNMDWRGRVYAVSMFNPDSPATTKGLLTLAKGKPIGKEGYYWLKIHGANCAGVDKVPFPERIKFIEENHENIMAC
    AKSPLENTWWAEQDSPFCFLAFCFEYAGVQFGASGYNCSLPLAFDGSCSGIQHFSAMLRDEVGGRAVNLLPSETVQDIYG
    IVAKKVNEILQADAINGTDNEVVTVTDENTGEISEKVKELIDKLAGQWLAYGVTRSVTKRSVMTLAYGSKEFGFRQQVLEDTI
    QPAIDSGKGLMFTQPNQAAGYMAKLIWESVSVTVVAAVEAMNWSSLRAKLLAAEVKDKKTGEILRKRCAVHWVTPDGFPV
    WQEYKKPIQTRLNLMFLGQFRLQPTINTNKDSEIDAHKQESGIAPNFVHSQLVALARKTVVWAHEKYGIESFALIHDSFGTIP
    ADAANLFKAVRETMVDTYESCDVLADFYDQFADQLHESQLDKMPALPAKGNLNLRDILESDVSEVV

    4. MNTINIAKNDFSDIELAAIPFNTLADHYGERLAREQLALEHESYEMGEARFRKMFERQLKAGEVADNAAAKPLITTLLPKMID
    AGIVEWFEEVKAKRGKRPTAFQFLQEIKPEAVAYITIKTTLACLTSADNTTVQAVASAIGRAIEDEARFGRIRDLEAKHFKKNVE
    EQLKEAQKHVYKKAFMQVVEADMLSKGLLGGEAWSSWHKEDSIHVGVRCIEMLIESTGMVSLHRQNAGVVGQDSETIELA
    PEYAEAIATRALLLRAISPMFQPCVVPPKPWTGITGGGYWANGRRPLALVRTHSKKALMRYEDVYMPEVYKAINIAQNTAWK
    INKKVLAVANVITKWKHGLLKGIPAIEREELPMKPEDIDMNPEALTAWKRAAAAVYRKDKARKSRRISLEFMLEQANKFANHK
    AIWFPYNMDWRGRVYAVSMFNPRTREFTKGLLTLAKGKPIGKEGYYWLKIHGANCAGVDKVPFPERIKFIEENHENIMACAK
    SPLENTWWAEQDSPFCFLAFCFEYAGVQIPIPKYNCSLPLAFDGSCSGIQHFSAMLRDEVGGRAVNLLPSETVQDIYGIVAK
    KVNEILQADAINGTDNEVVTVTDENTGEISEKVKEAKDALAGQWLAYGVTRSVTKRSVMTLAYGSKEFGFRQQVLEDTIQPA
    IDSGKGLMFTQPNQAAGYMAKLIWESVSVTVVAAVEAMNWRKSKAKLLAAEVKDKKTGEILRKRCAVHWVTPDGFPVWQE
    YKKPIQTRLNLMFLGQFRLQPTINTNKDSEIDAHKQESGIAPNFVHSQAAALARKTVVWAHEKYGIESFALIHDSFGTIPADAA
    NLFKAVRETMVDTYESCDVLADFYDQFADQLHESQLDKMPALPAKGNLNLRDILESDGADGL

    5. MNTINIAKNDFSDIELAAIPFNTLADHYGERLAREQLALEHESYEMGEARFRKMFERQLKAGEVADNAAAKPLITTLLPKMIK
    ELSKKWFEEVKAKRGKRPTAFQFLQEIKPEAVAYITIKTTLACLTSADNTTVQAVASAIGRAIEDEARFGRIRDLEAKHFKKNV
    EEQLRALGAHVYKKAFMQVVEADMLSKGLLGGEAWSSWHKEDSIHVGVRCIEMLIESTGMVSLHRQNAGVVGQDSETIEL
    APEYAEAIATRAAAIVRISPMFQPCVVPPKPWTGITGGGYWANGRRPLALVRTHSKKALMRYEDVYMPEVYKAINIAQNTAW
    KINKKVLAVANVITKWKHLSPLIIPAIEREELPMKPEDIDMNPEALTAWKRAAAAVYRKDKARKSRRISLEFMLEQANKFANHK
    AIWFPYNMDWRGRVYAVSMFNPSPELRTKGLLTLAKGKPIGKEGYYWLKIHGANCAGVDKVPFPERIKFIEENHENIMACAK
    SPLENTWWAEQDSPFCFLAFCFEYAGVQGALKAYNCSLPLAFDGSCSGIQHFSAMLRDEVGGRAVNLLPSETVQDIYGIVA
    KKVNEILQADAINGTDNEVVTVTDENTGEISEKVKDADIILAGQWLAYGVTRSVTKRSVMTLAYGSKEFGFRQQVLEDTIQPA
    IDSGKGLMFTQPNQAAGYMAKLIWESVSVTVVAAVEAMNWLSLRSKLLAAEVKDKKTGEILRKRCAVHWVTPDGFPVWQE
    YKKPIQTRLNLMFLGQFRLQPTINTNKDSEIDAHKQESGIAPNFVHSQSLSSARKTVVWAHEKYGIESFALIHDSFGTIPADAA
    NLFKAVRETMVDTYESCDVLADFYDQFADQLHESQLDKMPALPAKGNLNLRDILESDDELAA

    3. Produce Alternate Protein Models

    Using these alternate sequences, we could now activate the final component of our plan. The program built was designed to use the sequences generated to predict protein structure. The code was also able to conduct a docking process and perform molecular dynamics simulation (MD) to simulate a 5 ns dynamic interaction between protein and targeted DNA sequence in a 1-nm size water cube.

    Fig 1: End-results of protein models for 5 alternate RNA sequences generated.