diff --git a/wiki/pages/model.html b/wiki/pages/model.html
index b25d3ccfcd50ff9baf913c9d5e04730dad070120..af41435c486255ab40231fb2a6011c2c09132e7d 100644
--- a/wiki/pages/model.html
+++ b/wiki/pages/model.html
@@ -55,7 +55,7 @@ figcaption {
     <h4 style = "color:rgb(255, 255, 255)">Model of our engineered sequence is consistent with prior measurements of a disulfide bond</h4>
     <p>
       While the AlphaFold <a href="#ref1"><sup>[1]</sup></a> predictions are low confidence of the H-fibroin structure as a whole, the use of energy minimization and molecular 
-      dynamics <a href="#ref2"><sup>[2]</sup></a>,<a href="#ref3"><sup>[4]</sup></a>,<a href="#ref4"><sup>[7]</sup></a> builds confidence in the quality of this output. Most importantly, our model suggests that key residue-specific interactions 
+      dynamics <a href="#ref2"><sup>[2]</sup></a>,<a href="#ref3"><sup>[3]</sup></a>,<a href="#ref4"><sup>[4]</sup></a> builds confidence in the quality of this output. Most importantly, our model suggests that key residue-specific interactions 
       between L- and H-fibroin are preserved in our engineered sequences, relative to their native counterparts. A strong cysteine-cysteine disulfide 
       interaction between h- and L-fibroin is predicted to be preserved—for instance, the model suggests that the cysteine ~20 residues from the end 
       of our engineered H-fibroin and the cysteine between residues 150 and 200 of the L-fibroin are consistent with their expected relative positions