The Epic Collaboration: DeepMind and AMB LEO EBI Unveil Groundbreaking Alpha Fold Protein Structure Database
In July 2021, a historic collaboration between DeepMind and AMB LEO EBI was announced, marking a significant breakthrough in the field of protein structure prediction. This partnership has yielded an unprecedented achievement - the creation of a comprehensive database, known as the AlphaFold protein structure database, which covers approximately 98.5% of all human proteins, translating to around 20,000 to 30,000 proteins.
The AlphaFold database is a masterpiece of AI-driven research, leveraging cutting-edge machine learning algorithms to predict the three-dimensional structures of proteins. This monumental undertaking has far-reaching implications for researchers and scientists worldwide, offering a wealth of opportunities for designing new drugs, improving enzyme efficiency in industrial applications, and advancing our understanding of complex biological systems.
The AlphaFold database is an open platform, freely accessible to anyone, allowing researchers to explore and utilize the predicted protein structures. To access the database, one can simply type in the name of their desired protein and click on "Search." For example, attempting to find information on "Cytochrome" yields a plethora of results, including examples, PDB files, and detailed information about each protein.
Upon clicking on a particular protein, such as Cytochrome, users can access the predicted structure, which is displayed in three-dimensional format. This feature allows researchers to visualize the protein's shape, zoom in, and rotate it to gain a deeper understanding of its structure. The database also provides color-coded confidence levels for each prediction, with high-confidence structures represented by dark blue and lower-confidence structures marked with red.
The AlphaFold database is not only remarkable for its scope but also its accuracy. Most proteins have been predicted with high confidence, with the exception of loop regions, which are notoriously difficult to predict due to their unstructured nature. However, even these challenging regions can be accessed and explored within the database.
In addition to the comprehensive coverage of human proteins, the AlphaFold database will soon expand to include over 100 million proteins from the Uniref 90 database, providing a vast resource for researchers working with organisms beyond humans. This expansion is expected to occur in the coming months, marking another significant milestone in the development of this groundbreaking database.
For those interested in utilizing the AlphaFold database for their research, users can now enter their own protein sequences and test the predictive capabilities of the algorithm. This feature offers a unique opportunity for researchers and graduate students to engage with the data, design new proteins, and perform mutagenesis experiments to further refine the predictions.
The collaboration between DeepMind and AMB LEO EBI has sparked excitement among the scientific community, with many esteemed Nobel laureates in chemistry, physiology, medicine, and other fields expressing their enthusiasm for this achievement. Their endorsements validate the significance of this breakthrough, highlighting its potential to revolutionize various fields of research and applications.
The AlphaFold protein structure database serves as a testament to human ingenuity and the power of collaboration between industry leaders and scientific institutions. As researchers continue to explore and refine this groundbreaking resource, it is clear that the future of protein structure prediction holds immense promise for advancing our understanding of complex biological systems and developing innovative solutions for some of humanity's most pressing challenges.
For those interested in exploring the AlphaFold database further, links to relevant resources are provided in the video description. Supporting the channel by liking, subscribing, and enabling notifications will ensure that future content continues to be developed and shared with the community. The best way to learn data science is through hands-on experience; we invite you to embark on this journey of discovery and explore the vast possibilities offered by the AlphaFold protein structure database.
"WEBVTTKind: captionsLanguage: enin a recent video i was talking about the release of the alpha photo 2 and how deepmind shared their source code on github where other researchers and also other enthusiasts could get access to their code in order to build their very own predicted protein structure from the amino acid sequence and so in this video we're going to talk about an exciting collaboration between deepmind and also the ambo ebi which is a reputable bioinformatics institute and how that collaboration led to the release of the alpha fold protein structure database and so let's get started so in the previous video i briefly mentioned about this github of alpha photo 2 where you could get access to the code as a doctor and also they provided with some information on how you could download the necessary databases however if your computer is short on storage and you need a quicker way to test out alpha fold let me show you this google collab so this is the alpha photo 2 google colab and it is actually a tuned down version of the photo version of alpha fold and so in this tuned down version it does not make use of templates protein structure and therefore they mention here that the accuracy might suffer however for most proteins they're going to be pretty much near identical performance and so if you would like to give apple photo 2 a try you could check out this particular google collab and so i have already run this to cells to install the third-party software and the alpha fold executable version and here i'm inputting the example sequence that came along with the google colab and then click on the run and currently right now it's doing a search against the databases and then it's going to perform the prediction and allow me to download the prediction and so i'll provide you the link to this particular google collab and so you could give it a try so this is an awesome way to take alpha photo 2 for a test drive and so one exciting news about the partnership between the ambo ebi and the deepmind is that the alpha photo is being applied to make predictions on the entire human proteome which they have already done and it covers about 98.5 of all human proteins which is about 20 000 to 30 000 proteins and so the work described here is hosted on this particular database the alpha fold protein structure database and so if you would like to read some more details you could check out this particular research article and i'll provide you the link to it as well and so let's jump over to the news of the collaboration between deepmind and the amble ebi so this is the news about the epic collaboration which was released on 22nd of july 2021 and today's 23rd and so this is a recent news and i'm particularly very excited about this because the thing is all of the human proteins are now being predicted and they're offered publicly for anyone to get access to the protein structure and the exciting news is because that the ability to have access to these protein structures would be useful for researchers and that they could use it for designing new drugs they could use it to design better enzymes for industrial applications and so the possibilities are endless so i'll provide you the link to this particular blog post as well and let's take a quick look at some of the feedback from the scientific community and so you can see here that most of these are reputable nobel laureates in chemistry in physiology medicine and also in chemistry in 2009 as well and so all of them are very happy about this epic progress that deep mind has made and also the collaboration between deepmind and the ambo ebi and so let's have a look at the alpha fold protein structure database so here you could type in the protein of your interest and then click on search so why don't we do that let's type in let's see cytochrome and then it'll give you some examples and then let's click on the first one as as an example right very nice so it has the pdb file here which is the three-dimensional protein structure file format so you could click on it download it to your computer and then you could visualize the protein or you could use the web version here you could zoom in as well zoom out rotate it you could click on particular positions and it will highlight for you to see and so this is the zoomed in version very nice so they even color code the confidence that they have in the prediction so the high confidence will be in the dark blue while red color here will be lower confidence so most of the protein structure are pretty much in high confidence you can see like bluish color and yellow which is moderately good okay aside from the loop region here is in low confidence mode here and this is normal because loop regions are pretty unstructured and therefore they're more difficult to predict all right so this is a look at the page of the particular protein of interest let's go back to the front page and so they also provide you some other example as well let's say that i click on the e coli all right and then you can see that there are a couple of uncharacterized proteins that they have already made predictions for and so scientists who are working with some of these proteins and they haven't yet obtained the x-ray crystallographic structure of it they could take a glance at the predicted version here let's click on one of them so this is the alpha helix so it's a short segment so a small peptide and let's have a look at some of the information provided here on the front page so yeah right here another interesting point here is that they're mentioning here that in the coming months the alpha fo database will contain an expanded range of more than 100 million proteins from the uniref 90 database from uniprot and uniprot is a protein repository so there are over 100 million proteins spanning all organisms and so it will be interesting and it is definitely a big scientific breakthrough that has never been done before which is to perform a massive prediction of all 100 million proteins that are ever existed and so i'll keep you guys updated on that and so let's have a look at the prediction so it's still ongoing and it might take some time and you could also feel free to enter your own protein sequence here so if you're a researcher or a graduate student and you're working on designing some proteins engineering some proteins you could definitely perform some mutagenesis and put in your query protein here and take it for a test drive so i'll provide you the links in the video description and so if you find value in this video please support the channel by smashing the like button subscribing if you haven't already and also hit on the notification bell so that you will be notified of the next video and as always the best way to learn data science is to do data science and please enjoy the journeyin a recent video i was talking about the release of the alpha photo 2 and how deepmind shared their source code on github where other researchers and also other enthusiasts could get access to their code in order to build their very own predicted protein structure from the amino acid sequence and so in this video we're going to talk about an exciting collaboration between deepmind and also the ambo ebi which is a reputable bioinformatics institute and how that collaboration led to the release of the alpha fold protein structure database and so let's get started so in the previous video i briefly mentioned about this github of alpha photo 2 where you could get access to the code as a doctor and also they provided with some information on how you could download the necessary databases however if your computer is short on storage and you need a quicker way to test out alpha fold let me show you this google collab so this is the alpha photo 2 google colab and it is actually a tuned down version of the photo version of alpha fold and so in this tuned down version it does not make use of templates protein structure and therefore they mention here that the accuracy might suffer however for most proteins they're going to be pretty much near identical performance and so if you would like to give apple photo 2 a try you could check out this particular google collab and so i have already run this to cells to install the third-party software and the alpha fold executable version and here i'm inputting the example sequence that came along with the google colab and then click on the run and currently right now it's doing a search against the databases and then it's going to perform the prediction and allow me to download the prediction and so i'll provide you the link to this particular google collab and so you could give it a try so this is an awesome way to take alpha photo 2 for a test drive and so one exciting news about the partnership between the ambo ebi and the deepmind is that the alpha photo is being applied to make predictions on the entire human proteome which they have already done and it covers about 98.5 of all human proteins which is about 20 000 to 30 000 proteins and so the work described here is hosted on this particular database the alpha fold protein structure database and so if you would like to read some more details you could check out this particular research article and i'll provide you the link to it as well and so let's jump over to the news of the collaboration between deepmind and the amble ebi so this is the news about the epic collaboration which was released on 22nd of july 2021 and today's 23rd and so this is a recent news and i'm particularly very excited about this because the thing is all of the human proteins are now being predicted and they're offered publicly for anyone to get access to the protein structure and the exciting news is because that the ability to have access to these protein structures would be useful for researchers and that they could use it for designing new drugs they could use it to design better enzymes for industrial applications and so the possibilities are endless so i'll provide you the link to this particular blog post as well and let's take a quick look at some of the feedback from the scientific community and so you can see here that most of these are reputable nobel laureates in chemistry in physiology medicine and also in chemistry in 2009 as well and so all of them are very happy about this epic progress that deep mind has made and also the collaboration between deepmind and the ambo ebi and so let's have a look at the alpha fold protein structure database so here you could type in the protein of your interest and then click on search so why don't we do that let's type in let's see cytochrome and then it'll give you some examples and then let's click on the first one as as an example right very nice so it has the pdb file here which is the three-dimensional protein structure file format so you could click on it download it to your computer and then you could visualize the protein or you could use the web version here you could zoom in as well zoom out rotate it you could click on particular positions and it will highlight for you to see and so this is the zoomed in version very nice so they even color code the confidence that they have in the prediction so the high confidence will be in the dark blue while red color here will be lower confidence so most of the protein structure are pretty much in high confidence you can see like bluish color and yellow which is moderately good okay aside from the loop region here is in low confidence mode here and this is normal because loop regions are pretty unstructured and therefore they're more difficult to predict all right so this is a look at the page of the particular protein of interest let's go back to the front page and so they also provide you some other example as well let's say that i click on the e coli all right and then you can see that there are a couple of uncharacterized proteins that they have already made predictions for and so scientists who are working with some of these proteins and they haven't yet obtained the x-ray crystallographic structure of it they could take a glance at the predicted version here let's click on one of them so this is the alpha helix so it's a short segment so a small peptide and let's have a look at some of the information provided here on the front page so yeah right here another interesting point here is that they're mentioning here that in the coming months the alpha fo database will contain an expanded range of more than 100 million proteins from the uniref 90 database from uniprot and uniprot is a protein repository so there are over 100 million proteins spanning all organisms and so it will be interesting and it is definitely a big scientific breakthrough that has never been done before which is to perform a massive prediction of all 100 million proteins that are ever existed and so i'll keep you guys updated on that and so let's have a look at the prediction so it's still ongoing and it might take some time and you could also feel free to enter your own protein sequence here so if you're a researcher or a graduate student and you're working on designing some proteins engineering some proteins you could definitely perform some mutagenesis and put in your query protein here and take it for a test drive so i'll provide you the links in the video description and so if you find value in this video please support the channel by smashing the like button subscribing if you haven't already and also hit on the notification bell so that you will be notified of the next video and as always the best way to learn data science is to do data science and please enjoy the journey\n"
