hello everyone my name is ashak Ahmed and I work in the field of structure biology and bioinformatics today I'm going to introduce you K one decoding the molecular interactions of life so we will talk about kai1 we will discuss some results comparisons of K1 with Alpha fold in other currently running programs and I will also teach you how you can interpret the results received from 51 so let's begin first of all this slide is about the introduction of K1 K1 one actually explains the versin number so kai1 is introduced or released by Kai discovery and it is AI package or AI Trend package it is using AI Trend models and with the help of tai1 you can model protein protein interactions protein structures protein DNA and RNA interactions and above all you can model routine Lian predictions or protein Lian modeling or whatever you can also say a kind of protein Lian docking I'm presenting you this program and in this series this is the third program I'm presenting to you I have also taught you about the alpha fold three I have also taught you about the bolds one and now I'm going to teach you Kyan because Kyan claims in the results that it is producing better result than Alpha 43 in terms of protein and Lian POS predictions this SL is from their Pepper or from the website so the left side is Lian POS Buster set for Alpha 4 3 and this is ky and Rosita tt4 so Ros ttf f is far behind but tai1 is slightly better uh if we if you can see it's slightly bitter than Alpha 4 3 in Lian POS predictions are predicting the best poses for a protein and Lian modeling on the other hand they have also tested um for all atoms the kai1 single sequence and the K1 and the alpha fold 2. 3 and Alpha fold 3 the alpha fold 3 is tested here and for this range or for this uh panel they have tested the alpha fold 2. 3 version uh I think at that time uh Alpha F 3 was not releas so this could be the one reason but for all protein interaction uh it is better than Alpha 4 2.
3 for antibody protein prediction Kyan is once again better and for protein monomer predictions they are uh it's um not very significant but slightly B than Alpha 2 so these were all the the things so I believe as uh according to the results pi1 is a bit better or slightly better than the alpha 4 3 so you are allowed to test it today I'm going to teach you how can you use Kwan and then the the last slide we will teach at the end it's about how to interpret Kwan results so let's start and first learn how we can use Kyan for or uh let's say today I'm going to teach you protein Lian prediction okay so let's start so first of all you have to open your browser and write Kai Discovery server I will also drop this link in the description section for direct clicking but if you want if you're following this video so you can click the kai Discovery server and you will have this site lab. kaiis discovery. com/ dashboard so you will see three tabs Ms a tab template tab it's still disabled they will enable soon and specify restraints so if you are using MSA T it means you are providing multiple sequence alignment produced by you but if you are not using the MSA T then right now you can use the specify restrain Tab It Is by default selected so it is using a language model and if you are interested in some restraints you can okay so the first thing is to name a job model let's say I'm naming in test one okay now we are going to model a protein system with liin so I have opened the uni server and here I will write NDM One metalo beta leum as is okay I will select this structure just by default and here I will select uh the sequence and from this window I can select the popit the sequence now please pay attention you should see protein here now CLI here uh past here so now you have pasted your sequence okay now you can see copies in case you want to predict a dimer Trier tetramer etc etc then you can increase that number and subsequently in chain IDs you can see a b c d it means I I'm going for a tetramic protein if now I'm going for diic but in this case I'm going for a monomer and because it contains the signal peptide sequence so I have to remove the first uh 42 residues like EG EtG these are the signal peptide sequence so I am going just to remove EtG in your case you have to deal your sequence by yourself now I I'm also interested to model uh a Lian with it so add molecule now I I will open pupkin and beta liases or the NDM one contains two zinc onon so I type zinc here and in isomeric smilees you can see here so just copy this and click here Li in and say control V but I'm interested in two copies so it's here now the third thing I'm interested to include any drug I would say so I'm interested in penciling so from here you have to copy the isomeric smilees yes did it contrl C copy now test it it's a invalid corrector so I will click here and what's okay because I have selected protein here so that's why now it's fine so our system is ready so let's try tai1 and our system contains one liin penine twoing ions and should be modeled in the pocket and a protein sequence predict and I will name it the name of this run will be test one testore one and predict 3D structure job submitted successfully now you can click view okay these this is my job protein Li in Li in and it's it's saying it's processing and now we can refresh this and it it may take two to 3 minutes uh possibly so by that time I will download the crystal structure structure from the pdb for alignment or for the comparison purposes so let me download rcsb because we have also to check the the efficiency of ky1 for EXs I Know download pdb format okay so it's here we will open it from the download folder once we receive our kai1 results still process so I will pause this video for a while and until this process is complete uh I think it's done with okay success now you can see this is our structure and you can also see there are two zinc ion let me check you can see here there are two zinc ions and they are in a perfect position because they have engage the catalytic Triad of uh nm1 so this is zinc one and this is zinc 2 and where is our Lian so here is our Lian you can see okay now how much good is this structure this is important so this this is the iptm and PTM so first of all we are going to to discuss this things these metrices how you will understand like your docking or your modeling is near to the real structure or it's just a wrong uh wrong modeling or a wrong docking okay so from here you can see majority of the things are in wide Zone and the wide zone is near to zero and this is the expected position error so it means from kaiwan uh these this prediction is Top Class PTM is96 means accuracy is 96% but to understand the PTM and I PTM uh we have to discuss this slide shortly okay there is a term it's term name is TM TM means template modeling the the TM score means the template modeling score give given to you from the program so what is PTM predicted template modeling score remember in our case we received this almost 0.
9 6 it means 96% so what the ptms tells you if the PTM score is above 0. 5 it means that you can consider this structure okay and in our case it is96 so it means that the kaian recommends the highest accuracy I will use the world highest the highest accuracy for the structure to be used okay but in our case we have another portion that is iptm why we have iptm because we are using two different things if you are modeling only a protein structure a single protein structure you may not receive I PTM but if you are dealing with two different proteins or if you are dealing with the protein and DNA or the protein and Lian then you will receive the iptm score so the iptm score actually me the accuracy of the predicted position of the two interaction interacting Partners it can be protein protein complex it can be protein l in complex blah blah blah the higher value is the better one but if your value is higher than point8 it means be confident and go ahead if your value belongs to 6 and8 it means this is a gr Zone grone means that in some cases it can be wrong but it can be right also it's a kind of this is what you call it a gry Zone but in our case we got this almost 0. 75 and something let me check again uh 0.
75 years here so it means it is in the gray Zone but I personally recommend if it is above 7 you can consider that gr Zone towards the true positive so this the you need to understand these things because only modeling is not um the main goal like you you use Alpha fold three or you use boards or you use IP uh this one the kai one and you thought that okay done no if these scores are on the lower side then you have to be careful for publishing or for accepting is the result is a true positive okay now finally we are going to download the results so download results and I will receive the results here test 1.
