01

docs/Texts/0/01.md

@@ -0,0 +1,84 @@
+# The Question!
+---
+##### Semi-Private | S.Alireza Hashemi
+
+<div style="text-align: justify"> 
+After the first-round results were announced, we met with (@Kave_nasr) to tackle the following question: Could we establish a single metric—or a combination of metrics—that reliably indicates binding confidence among the top-ranked candidates?
+Drawing on data from the IL-7a case study and all 201 sequences from the first round, we predicted each structure, took the top-ranked (rank 1) model, and then applied the Rosetta Score function and interface-analyzer (as used by Vasquez and Kuhlman) to seek meaningful correlations. (All structures underwent the fast-relax protocol, in parallel with AF2, to gain some computational advantages over AF2 relaxation.)
+Below are a few of the plots exploring an assumed correlation between iPAE and X, where X includes:
+</div>
+
+•	Interface energy
+
+•	Hbond energy
+
+•	Interface SASA
+
+•	Number of Interface Residues
+
+•	Complexed Energy
+
+•	Complexed SASA
+
+•	Side_Score
+
+•	Hydrophobic Interactions
+
+•	Electrostatic_Score
+
+•	Fold_Tree
+
+•	Rotamers_Score
+
+
+---
+
+IL7a Case:
+
+    •	Baker (B/U): 22/10
+
+    •	Adaptyv (B/U): 18/14
+
+#### --> [📄 Open [IL7a] PDF in New Tab](./IL7A.pdf){:target="_blank"}
+
+![pdf](./01.png){ align=left style="height:;width:550px; margin-top: 0px;" }
+
+
+<div style="height: 250px;"></div>
+<br><br>
+
+
+<div style="text-align: justify">
+And as the same procedure for the EGFR case:
+</div>
+
+---
+
+EGFR (Round1) Case:
+
+    •	201 Protein:
+    •	Expressed: 147
+    •	(B/U): 7/147
+
+#### --> [📄 Open [EGFR] PDF in New Tab](./EGFR.pdf){:target="_blank"}
+
+---
+### Discussion
+<div style="text-align: justify"> 
+
+For now, we plan to include other bakers’ work from RFDiffusion and Round 2 of Adaptyv bio, as it helps clarify noise and, importantly, addresses two main points: first, reducing the bias from the large number of successful proteins reported in the RFDiffusion paper, and second, classifying binders by the model or other shared criteria.
+
+For example, if we look at Vázquez’s work and the recommended filters (ddG < -40 & pAE <10), it does help screen out cases that don’t meet those conditions. However, there are two issues:
+</div>
+
+• Although it effectively filters non-binders, it does not reliably confirm binding confidence.
+
+• For semi-natural binders disqualified in Round 1, these filters did not align with the results.
+
+<div style="text-align: justify"> 
+It’s also worth noting that we used highly selective IL-7a data. We are now calculating Round 2 metrics and incorporating more RFDiffusion data to achieve better classification.
+</div>
+
+### Deviation from correlation observed in natural binders:
+
+![Correlation](./Correlation.png)

mkdocs.yml

@@ -77,6 +77,7 @@ extra_javascript:
 
 plugins:
   - search
+
 
 nav:
   - Home: ./index.md