 Dear students, now I will introduce you to a very interesting algorithm for structure prediction. This algorithm is called the 3D1D Bowie algorithm and we will see how it differs from homology modeling, ab initio and fold recognition. So first of all, if you remember, homology modeling was employed for predicting the structure of a protein by using its primary sequence that is the amino acid sequence. If we had a high match or high alignment between the sequence and the sequence from the PDB database. Now there was a case when we had low alignment between the sequence that we want to create a structure from and the PDB. So in that case we employed the fold recognition or threading. Now towards the 3D1D Bowie algorithm, so this was proposed by Bowie et al in 1991. So what it does is, it converts the 3D structural profile of a protein structure into a 1D array and then you can compare this 1D array to all the amino acids within the sequence for which you want to predict the structure. So if you get a very good match, then you can declare that structure and successfully predict it. So to start with, you align the target that is the sequence to the 1D profile. So let's see how it works. So towards the 3D1D algorithm, the inputs and outputs are as follows. So the first thing that you need to do is identify the amino acids that are there in the protein core. So as you know, the protein has a core, a hydrophobic core, hydrophobic core and this is the external structures. So in this core, you have to identify the amino acids, okay? So this is step number one. Then secondly, you have to identify the side chain positioning. So the side chains for each amino acid are considered and lastly the solubility of the surface amino acids. So considering these three things and some other properties which in case of the 3D1D were six, so you create a profile. So the profile can be, for instance, let's say for amino acid alanine. So does it occur in the protein core? Let's say yes. So one side chain positioning, let's say arbitrarily in Rotamer configuration one and solubility let's say zero and so on. So this is one profile for alanine. Similarly you can create a profile for the 20 different amino acids. Of course alanine may occur in a situation where it is not at the protein core and therefore may have a zero and it may have a different Rotamer configuration as well and its solubility may differ slightly depending on its location and its neighboring amino acids. So one amino acid can have multiple profiles. So the two take home messages are that you have to create a profile that is ones and zeros for each amino acid for each instance in a protein structure. Now if you are able to do that then you will have multiple profiles for each amino acid. Okay next you also count which amino acid occurs in which secondary structure. For instance in case of alpha helix and our example alanine A. So most probably it will be in the alpha helixes but sometimes it may so happen that it will also be in a beta sheet. So depending on how many times alanine occurs in alpha helix, beta sheet and loops you create a profile for alanine in the secondary structures as well. So if you create a profile for let's say six attributes and three secondary structures then you will have 18 different distinct states for each amino acid. So if you have 20 amino acids then 18 states per amino acid so therefore 18 multiplied by 20 so you will have about 360 different profiles. Remember that these profiles are one dimensional now or simply ones and zeros. So next you calculate the probability of an amino acid A occurring in an environment in J. So the profiles are also known as the environments. So in all you had 360 different profiles for 20 amino acids and 18 different profiles per amino acid. So you have to calculate the probability of finding a amino acid A in a profile J. Once you have computed that then you find out the probability of finding the same amino acid A anywhere in the sequence. Now using this very simple formula for computing the score. So you take a log of the ratio of probability A given J that is this one divided by probability of A which is given here. So if you take a log of this you will have a score for amino acid A in a given profile or environment J. All you have to do is to maximize the sum of these scores for the fold. So in the fold there will obviously be multiple amino acids. So let's say if you have 10 amino acids in a fold then you have to sum all these S A J's maximize them. So this will be the score that will be the profile used to predict the structure. So in conclusion the 3D 1D algorithm converts the structural and physical as well as chemical properties of an amino acid into a 1D profile and then maximizes the score of such profiles. Once you have computed the score for each amino acid within each profile you can predict the structure that you are interested in.