Vision of Pak TIK: AlphaFold and Biotech Revolution for Human Urban Evolution
Tauhid Nur Azhar
Many things in this mysterious life are often shrouded in questions, and finding the meaning of life itself is one of the things that become a major concern, and maintaining it is not just a matter of personal choice?
Pak T Indra Kesuma is a teacher who is very concerned about the development of artificial intelligence technology. Perhaps he can be considered a co-founder of this artificial intelligence technology or KORIKA, which is also actively involved in the formulation of the National Artificial Intelligence Strategy or Stranas AI.
One of the focuses of his concern is the implementation of AI in the field of biotechnology, healthcare, and health surveillance. More than just the advancement of Nobel Chemistry in 2024, awarded to Prof David Baker and the team of AI innovation pioneers in structural modeling. Where the birth of Nobel Chemistry in 2024 was inspired by the development of AlphaFold. David Baker was born in 1962 in Seattle, America (USA). He earned his doctorate in 1989 from the University of California, Berkeley. Baker is also known as a prominent professor at the University of Washington, Seattle, and a researcher at the Howard Hughes Medical Institute, USA.
Since the birth of Nobel Chemistry in 2024, it has inspired an innovation breakthrough that combines advanced artificial intelligence technology with molecular biology and organic chemistry.
This innovation is AlphaFold. What is AlphaFold? AlphaFold is a computational model for predicting and modeling protein structures that can identify and predict the structure of unknown molecules, which were previously unknown. Since its development by DeepMind, its technology company, in 3 versions. AlphaFold1, 2, and 3.
AlphaFold was developed by a team of researchers at DeepMind, a leading artificial intelligence company based in London and owned by Alphabet Inc. The AlphaFold project was led by a team of leading scientists, including John Jumper, who led the AlphaFold project and played a key role in developing the method of deep learning for protein structure prediction. There is also Richard Evans, a senior researcher at DeepMind focused on algorithm development and protein structure modeling in AlphaFold. Also, there is Andrew Sr, a researcher in the field of machine learning who played a significant role in improving the accuracy of protein structure prediction methods. Not to forget, Pushmeet Kohli, the head of the science department at DeepMind who provides strategic direction in various science projects, including AlphaFold.
Last but not least, there is Demis Hassabis, the CEO and one of the founders of DeepMind, who provides strategic direction and vision in the development of AlphaFold technology.
The development of AlphaFold also involves collaboration among experts from various disciplines, including bioinformatics, computational biology, and artificial intelligence. Together, they developed a deep learning-based system that uses transformer architectures to translate amino acid sequences into protein structures more accurately.
If we want to see and study the development of this era, it would be good for us to study the evolution of AlphaFold starting from version 1. Where AlphaFold 1 is the initial model developed by DeepMind to predict protein structure. AlphaFold 1 follows the guidelines of the Critical Assessment of Techniques for Protein Structure Prediction (CASP13) in 2018 and achieved the highest score among other methods.
The model uses machine learning techniques based on Convolutional Neural Networks (CNN) to predict the spatial arrangement of amino acids in a protein chain and the resulting structure. AlphaFold 1 creates a distance map that shows the spatial distance between each residue pair. Despite this innovation, AlphaFold 1’s predictions still have limitations in terms of accuracy and efficiency in predicting complex structures.
The next more up-to-date version in terms of technology and processing capabilities is AlphaFold2. Where AlphaFold 2 represents a significant advancement from the first version, released in 2020. This model improves the signaling from the first version, which was released in 2020. The model uses a more advanced architecture, namely Transformer Neural Networks, which can capture complex interactions between amino acid residues in a protein. In CASP14, AlphaFold 2 demonstrated superior performance with accuracy that surpasses experimental methods, such as X-ray crystallography.
AlphaFold 2 uses Multiple Sequence Alignment (MSA) and co-evolution techniques to infer the spatial arrangement of amino acids, providing a deeper understanding of the interactions between residues in a three-dimensional structure. One of the main advantages of AlphaFold 2 is its ability to predict protein structures with high precision in a shorter time, making it a valuable tool in biological research and medicine.
Additionally, AlphaFold 2 does not solely rely on distance maps to infer spatial distances between residues. The model also considers additional data from various species to improve the identification of evolutionary patterns in proteins.
The model AI analyzes amino acid sequences from different species to observe how amino acids interact within the three-dimensional structure, using the principle of co-evolution, where amino acids that interact within the protein structure co-evolve together. The iterative analysis process is carried out by a neural network (neural network) with a transformer, to identify key elements in the structure. Additional data from other protein structures, if available from the first stage.
AlphaFold2 generates a series of potential structures based on amino acid sequences and then refines them through an iterative process to produce the final structure. The AI model also calculates the probability of the accuracy of the generated structures compared to the actual structures.
Although still in the development and testing phase, AlphaFold version 3 is expected to improve the limitations and methods present in AlphaFold 2 by increasing the prediction of complex multi-protein interactions and how these proteins interact within the cell. This version is expected to better predict protein-protein interactions in the cellular environment.
Proteins are one of the biomolecules that have a three-dimensional structure. RNA molecules and RNA-protein complexes also have structures that function in many biological processes. AlphaFold 3 may be able to improve predictions on non-protein molecules, which are very beneficial for biological research and development of therapeutic RNA.
Many things in this mysterious life are often shrouded in questions, and finding the meaning of life itself is one of the things that become a major concern, and maintaining it is not just a matter of personal choice?
Pak T Indra Kesuma is a teacher who is very concerned about the development of artificial intelligence technology. Perhaps he can be considered a co-founder of this artificial intelligence technology or KORIKA, which is also actively involved in the formulation of the National Artificial Intelligence Strategy or Stranas AI.
One of the focuses of his concern is the implementation of AI in the field of biotechnology, healthcare, and health surveillance. More than just the advancement of Nobel Chemistry in 2024, awarded to Prof David Baker and the team of AI innovation pioneers in structural modeling. Where the birth of Nobel Chemistry in 2024 was inspired by the development of AlphaFold. David Baker was born in 1962 in Seattle, America (USA). He earned his doctorate in 1989 from the University of California, Berkeley. Baker is also known as a prominent professor at the University of Washington, Seattle, and a researcher at the Howard Hughes Medical Institute, USA.
Since the birth of Nobel Chemistry in 2024, it has inspired an innovation breakthrough that combines advanced artificial intelligence technology with molecular biology and organic chemistry.
This innovation is AlphaFold. What is AlphaFold? AlphaFold is a computational model for predicting and modeling protein structures that can identify and predict the structure of unknown molecules, which were previously unknown. Since its development by DeepMind, its technology company, in 3 versions. AlphaFold1, 2, and 3.
AlphaFold was developed by a team of researchers at DeepMind, a leading artificial intelligence company based in London and owned by Alphabet Inc. The AlphaFold project was led by a team of leading scientists, including John Jumper, who led the AlphaFold project and played a key role in developing the method of deep learning for protein structure prediction. There is also Richard Evans, a senior researcher at DeepMind focused on algorithm development and protein structure modeling in AlphaFold. Also, there is Andrew Sr, a researcher in the field of machine learning who played a significant role in improving the accuracy of protein structure prediction methods. Not to forget, Pushmeet Kohli, the head of the science department at DeepMind who provides strategic direction in various science projects, including AlphaFold.
Last but not least, there is Demis Hassabis, the CEO and one of the founders of DeepMind, who provides strategic direction and vision in the development of AlphaFold technology.
The development of AlphaFold also involves collaboration among experts from various disciplines, including bioinformatics, computational biology, and artificial intelligence. Together, they developed a deep learning-based system that uses transformer architectures to translate amino acid sequences into protein structures more accurately.
If we want to see and study the development of this era, it would be good for us to study the evolution of AlphaFold starting from version 1. Where AlphaFold 1 is the initial model developed by DeepMind to predict protein structure. AlphaFold 1 follows the guidelines of the Critical Assessment of Techniques for Protein Structure Prediction (CASP13) in 2018 and achieved the highest score among other methods.
The model uses machine learning techniques based on Convolutional Neural Networks (CNN) to predict the spatial arrangement of amino acids in a protein chain and the resulting structure. AlphaFold 1 creates a distance map that shows the spatial distance between each residue pair. Despite this innovation, AlphaFold 1’s predictions still have limitations in terms of accuracy and efficiency in predicting complex structures.
The next more up-to-date version in terms of technology and processing capabilities is AlphaFold2. Where AlphaFold 2 represents a significant advancement from the first version, released in 2020. This model improves the signaling from the first version, which was released in 2020. The model uses a more advanced architecture, namely Transformer Neural Networks, which can capture complex interactions between amino acid residues in a protein. In CASP14, AlphaFold 2 demonstrated superior performance with accuracy that surpasses experimental methods, such as X-ray crystallography.
AlphaFold 2 uses Multiple Sequence Alignment (MSA) and co-evolution techniques to infer the spatial arrangement of amino acids, providing a deeper understanding of the interactions between residues in a three-dimensional structure. One of the main advantages of AlphaFold 2 is its ability to predict protein structures with high precision in a shorter time, making it a valuable tool in biological research and medicine.
Additionally, AlphaFold 2 does not solely rely on distance maps to infer spatial distances between residues. The model also considers additional data from various species to improve the identification of evolutionary patterns in proteins.
The model AI analyzes amino acid sequences from different species to observe how amino acids interact within the three-dimensional structure, using the principle of co-evolution, where amino acids that interact within the protein structure co-evolve together. The iterative analysis process is carried out by a neural network (neural network) with a transformer, to identify key elements in the structure. Additional data from other protein structures, if available from the first stage.
AlphaFold2 generates a series of potential structures based on amino acid sequences and then refines them through an iterative process to produce the final structure. The AI model also calculates the probability of the accuracy of the generated structures compared to the actual structures.
Although still in the development and testing phase, AlphaFold version 3 is expected to improve the limitations and methods present in AlphaFold 2 by increasing the prediction of complex multi-protein interactions and how these proteins interact within the cell. This version is expected to better predict protein-protein interactions in the cellular environment.
Proteins are one of the biomolecules that have a three-dimensional structure. RNA molecules and RNA-protein complexes also have structures that function in many biological processes. AlphaFold 3 may be able to improve predictions on non-protein molecules, which are very beneficial for biological research and development of therapeutic RNA.
AlphaFold 3 is also expected to integrate dynamic molecular simulations (molecular dynamics) to better understand protein behavior under various conditions.
With a more efficient algorithm, AlphaFold 3 is expected to predict protein structures in a shorter time, making it applicable in industrial settings. This version is expected to better predict protein-protein interactions in the cellular environment.
Proteins are one of the biomolecules that have a three-dimensional structure. RNA molecules and RNA-protein complexes also have structures that function in many biological processes. AlphaFold 3 may be able to improve predictions on non-protein molecules, which are very beneficial for biological research and development of therapeutic RNA.
AlphaFold 3 is expected to integrate dynamic molecular simulations (molecular dynamics) to better understand protein behavior under various conditions.
With a more efficient algorithm, AlphaFold 3 is expected to predict protein structures in a shorter time, making it applicable in industrial settings. This version is expected to better predict protein-protein interactions in the cellular environment.
Proteins are one of the biomolecules that have a three-dimensional structure. RNA molecules and RNA-protein complexes also have structures that function in many biological processes. AlphaFold 3 may be able to improve predictions on non-protein molecules, which are very beneficial for biological research and development of therapeutic RNA.
AlphaFold 3 is expected to integrate dynamic molecular simulations (molecular dynamics) to better understand protein behavior under various conditions.
With a more efficient algorithm, AlphaFold 3 is expected to predict protein structures in a shorter time, making it applicable in industrial settings. This version is expected to better predict protein-protein interactions in the cellular environment.
Proteins are one of the biomolecules that have a three-dimensional structure. RNA molecules and RNA-protein complexes also have structures that function in many biological processes. AlphaFold 3 may be able to improve predictions on non-protein molecules, which are very beneficial for biological research and development of therapeutic RNA.
AlphaFold 3 is expected to integrate dynamic molecular simulations (molecular dynamics) to better understand protein behavior under various conditions.
With a more efficient algorithm, AlphaFold 3 is expected to predict protein structures in a shorter time, making it applicable in industrial settings. This version is expected to better predict protein-protein interactions in the cellular environment.
Proteins are one of the biomolecules that have a three-dimensional structure. RNA molecules and RNA-protein complexes also have structures that function in many biological processes. AlphaFold 3 may be able to improve predictions on non-protein molecules, which are very beneficial for biological research and development of therapeutic RNA.
AlphaFold 3 is expected to integrate dynamic molecular simulations (molecular dynamics) to better understand protein behavior under various conditions.
With a more efficient algorithm, AlphaFold 3 is expected to predict protein structures in a shorter time, making it applicable in industrial settings. This version is expected to better predict protein-protein interactions in the cellular environment.
Proteins are one of the biomolecules that have a three-dimensional structure. RNA molecules and RNA-protein complexes also have structures that function in many biological processes. AlphaFold 3 may be able to improve predictions on non-protein molecules, which are very beneficial for biological research and development of therapeutic RNA.
AlphaFold 3 is expected to integrate dynamic molecular simulations (molecular dynamics) to better understand protein behavior under various conditions.
With a more efficient algorithm, AlphaFold 3 is expected to predict protein structures in a shorter time, making it applicable in industrial settings. This version is expected to better predict protein-protein interactions in the cellular environment.
Further Reading
1. AlQuraishi, M. (2019). End-to-end differentiable learning of protein structure. Cell Systems, 8(4), 292–301. https://doi.org/10.1016/j.cels.2019.03.006
2. Baek, M., DiMaio, F., Anishchenko, I., Dauparas, J., Ovchinnikov, S., Lee, G. R., … & Baker, D. (2021). Accurate prediction of protein structures and interactions using a three-track neural network. Science, 373(6557), 871–876. https://doi.org/10.1126/science.abj8754
3. Bepler, T., & Berger, B. (2021). Learning the protein language: Evolution, structure, and function. Cell Systems, 12(6), 654–669. https://doi.org/10.1016/j.cels.2021.05.017
4. Berman, H. M., Battistuz, T., Bhat, T. N., Bluhm, W. F., Bourne, P. E., Burkhardt, K., … & Zardecki, C. (2002). The Protein Data Bank. Acta Crystallographica Section D: Biological Crystallography, 58(6), 899–907. https://doi.org/10.1107/S0907444902003451
5. Bhattacharya, D., & Cheng, J. (2017). DeNovo structure prediction using deep learning. Computational and Structural Biotechnology Journal, 15, 379–382. https://doi.org/10.1016/j.csbj.2017.03.006
6. Callaway, E. (2020). “It will change everything”: DeepMind’s AI makes gigantic leap in solving protein structures. Nature, 588(7837), 203–204. https://doi.org/10.1038/d41586-020-03348-4
7. Dill, K. A., & MacCallum, J. L. (2012). The protein-folding problem, 50 years on. Science, 338(6110), 1042–1046. https://doi.org/10.1126/science.1219021
8. Evans, R., & Jumper, J. (2020). AlphaFold’s success at CASP13. Proteins: Structure, Function, and Bioinformatics, 88(9), 1222–1227. https://doi.org/10.1002/prot.25909
9. Jumper, J., Evans, R., Pritzel, A., Green, T., Figurnov, M., Tunyasuvunakool, K., … & Hassabis, D. (2021). Highly accurate protein structure prediction with AlphaFold. Nature, 596(7873), 583–589. https://doi.org/10.1038/s41586-021-03819-2
10. Kryshtafovych, A., Schwede, T., Topf, M., Fidelis, K., & Moult, J. (2021). Critical assessment of methods of protein structure prediction (CASP) — Round XIII. Proteins: Structure, Function, and Bioinformatics, 89(12), 1687–1699. https://doi.org/10.1002/prot.26160
11. Marks, D. S., Colwell, L. J., Sheridan, R., Hopf, T. A., Pagnani, A., Zecchina, R., & Sander, C. (2011). Protein 3D structure computed from evolutionary sequence variation. PLoS One, 6(12), e28766. https://doi.org/10.1371/journal.pone.0028766
12. Mills, C. L., Garg, R., & Zhuang, W. (2020). Machine learning for structural bioinformatics. Bioinformatics, 36(11), 3450–3462. https://doi.org/10.1093/bioinformatics/btaa235
13. Moult, J., Pedersen, J. T., Judson, R., & Fidelis, K. (1995). A large-scale experiment to assess protein structure prediction methods. Proteins: Structure, Function, and Genetics, 23(3), ii-v. https://doi.org/10.1002/prot.340230303
14. Ovchinnikov, S., Park, H., Kim, D. E., DiMaio, F., & Baker, D. (2018). Protein structure prediction using Rosetta in CASP12. Proteins: Structure, Function, and Bioinformatics, 86, 113–121. https://doi.org/10.1002/prot.25390
15. Pereira, J., Simpkin, A. J., Hartmann, M. D., Rigden, D. J., Keegan, R. M., & Lupas, A. N. (2021). High-accuracy protein structure prediction in CASP14. Proteins: Structure, Function, and Bioinformatics, 89(12), 1681–1686. https://doi.org/10.1002/prot.26171
16. Raifeld, M., & AlQuraishi, M. (2021). Machine learning in protein structure prediction. Annual Review of Biophysics, 50, 121–140. https://doi.org/10.1146/annurev-biophys-061220-094522
17. Rives, A., Meier, J., Sercu, T., Goyal, S., Lin, Z., Liu, J. Z., … & Kohli, P. (2021). Biological structure and function emerge from scaling unsupervised learning to 250 million protein sequences. Proceedings of the National Academy of Sciences, 118(15), e2016239118. https://doi.org/10.1073/pnas.2016239118
18. Röthlisberger, D., Khersonsky, O., Wollacott, A. M., Jiang, L., DeChancie, J., Betker, J., … & Baker, D. (2008). Kemp elimination catalysts by computational enzyme design. Nature, 453(7192), 190–195. https://doi.org/10.1038/nature06879
19. Seemayer, S., Gruber, M., & Söding, J. (2014). CCMpred — Fast and precise prediction of protein residue–residue contacts from correlated mutations. Bioinformatics, 30(21), 3128–3130. https://doi.org/10.1093/bioinformatics/btu500
20. Senior, A. W., Evans, R., Jumper, J., Kirkpatrick, J., Sifre, L., Green, T., … & Hassabis, D. (2020). Improved protein structure prediction using potentials from deep learning. Nature, 577(7792), 706–710. https://doi.org/10.1038/s41586-019-1923-7
21. Sercu, T., & Goel, S. (2016). Convolutional networks for protein-protein interaction site prediction. Bioinformatics, 32(18), i425-i434. https://doi.org/10.1093/bioinformatics/btw473
22. Skolnick, J., & Gao, M. (2019). Interplay of physics and evolution in the likely origin of protein biochemical function. Proceedings of the National Academy of Sciences, 116(35), 17591–17600. https://doi.org/10.1073/pnas.1907771116
23. Tunyasuvunakool, K., Adler, J., Wu, Z., Green, T., Zielinski, M., Žídek, A., … & Jumper, J. (2021). Highly accurate protein structure prediction for the human proteome. Nature, 596(7873), 590–596. https://doi.org/10.1038/s41586-021-03828-1
24. Wang, S., Sun, S., Li, Z., Zhang, R., & Xu, J. (2017). Accurate de novo prediction of protein contact map by ultra-deep learning model. PLoS Computational Biology, 13(1), e1005324. https://doi.org/10.1371/journal.pcbi.1005324
25. Xu, J. (2019). Distance-based protein folding powered by deep learning. Proceedings of the National Academy of Sciences, 116(35), 16856–16865. https://doi.org/10.1073/pnas.1821300116
