link AF3 to main page, rm irrelevant casp-capri

Author: AnnaKravchenko

education/molmod_online/CASP_CAPRI16.png

@@ -148,24 +148,11 @@ This naturally raises an important question: how accurate are the models produce
     If the structures appear very similar in cartoon representation, consider switching to a lines or sticks representation to inspect side-chain conformations more closely.
 </a>
 
-
 In structural modelling, the quality of the predicted models depends on many factors, one of the most important being the amount of relevant information available. 
-As you observed during homology modelling, there is extensive experimental information for both MDM2 and the p53–MDM2 complex, including experimentally solved structures of human MDM2 bound to p53 (PDB: 1YCR), which can serve as a modelling template. 
+As you observed during homology modelling, there is extensive experimental information for both MDM2 and the p53–MDM2 complex, including multiple experimentally solved structures of human MDM2 bound to p53 (PDB: 1YCR, 4HFZ etc.) which can serve as modelling templates. 
 In such cases, producing an accurate prediction is relatively straightforward.
 The situation is very different for more challenging targets, where experimental information is sparse or entirely absent. 
 A useful way to assess the current state of the field is through community-wide blind prediction experiments such as CASP and the Critical Assessment of Predicted Interactions ([CAPRI](https://www.capri-docking.org/){:target="_blank"}).
 In these challenges, participants are given only the sequences of the constituents of the complex, whose structure have been experimentally determined but not yet released, providing an objective benchmark for computational methods.
-
-Looking at the results of the most recent analyzed round (CASP16/CAPRI - 2024), we see that the AlphaFold Server participates but is not among the top-performing approaches. 
-In fact, several human teams consistently produce more accurate models. 
-This is likely because they combine AlphaFold-based predictions with other modelling tools and, crucially, expert human judgment - the same types of skills you applied when performing homology modelling of MDM2, analyzing peptide MD simulations, and evaluating docking models.
-
-<img src="education/molmod_online/CASP_CAPRI16.png" style="margin-left: auto;margin-right:auto">
-
-These results also highlight a broader challenge in structural modelling, scoring. The plot includes results from MassiveFold, a dataset of 8,040 models per target generated using AlphaFold2. 
-As shown by the “best-of-best” bar, a high-quality model is almost always present somewhere in this large set. 
-However, identifying that model is far from trivial. The gap between the “best-of-best” and “MassiveFold-best” bars illustrates how difficult it is, even for expert teams, to reliably select the best model from a the pool.
-This points to two key limitations of current approaches. Firstly, better scoring id needed to identify the best model out of the pool.
-Secondly, generating thousands of models per target requires substantial computational resources. 
-Ideally, future modelling tools will be able to produce a small number of high-quality models directly, reducing the need for exhaustive sampling.
-In short, while tools like AlphaFold3 represent a major advance, they do not eliminate the need for careful analysis, complementary methods, and human expertise, especially for challenging targets.
+The results of these challenges show that output of AlphaFold Server is often not as accurate as the output generated by the human teams, especially when it comes to the challenging targets.
+To sum up, while tools like AlphaFold3 represent a major advance, they do not eliminate the need for careful analysis, complementary methods, and human expertise. 

education/molmod_online/alphafold3.md

@@ -227,16 +227,16 @@ If time is an issue, start with the molecular dynamics simulations and, while th
 <table class="two-col-table">
   <tr>
     <td></td>
-    <td><center><i><b>AlphaFold 2</b></i></center></td>
+    <td><center><i><b>AlphaFold 3</b></i></center></td>
   </tr>
   <tr>
     <td style="padding-right: 20px" valign="top">
-      In this bonus module you will discover the power of artificial intelligence (AI) for structural biology. We will introduce AlphaFold 2 and use it to model the MDM2/p53 protein-peptide complex from sequence only.
+      In this bonus module you will discover the power of artificial intelligence (AI) for structural biology. We will introduce AlphaFold 3 and use it to model the MDM2/p53 protein-peptide complex from sequence only.
     </td>
     <td style="width: 220px">
-      <a href="/education/molmod_online/alphafold"
-        alt="Modelling the MDM2/p53 complex using AlphaFold."
-        title="Modelling the MDM2/p53 complex using AlphaFold.">
+      <a href="/education/molmod_online/alphafold3"
+        alt="Modelling the MDM2/p53 complex using AlphaFold3."
+        title="Modelling the MDM2/p53 complex using AlphaFold3.">
         <img src="/images/molmod/alphafold_link-cropped.png" class="col-table" style="border-radius: 50%;">
       </a>
     </td>