deletions | additions       diff --git a/Results_P_about_specificity_of__.tex b/Results_P_about_specificity_of__.tex  new file mode 100644  index 0000000..b7946bc  --- /dev/null  +++ b/Results_P_about_specificity_of__.tex 
...
Results  P about specificity of fragments  P about VAMS performance  P about FOMS performance  P about time difference if I can  P about correlation plot  General Observations:  Protein targets with ligands that have many key interactions for their binding mode did not perform as favorably as targets with ligands that have a few key interactions. This makes sense since target with relatively few interactions could have the important features for protein inhibition covered by the SMARTS expression and fragment pre-alignment. These targets could easily reject compounds that didn’t match the SMARTS expression key for understanding the mode of inhibition.   In general, FOMS performed better than VAMS for every protein target. This indicates that it is an effective pre-screen, especially since VAMS was shown to be a very effective pre-screen earlier (cite VAMS).   In general, more specific SMARTS expressions led to a higher performance for both FOMS and VAMS. This makes sense since higher specificity of fragment choice will lead to identifying only the most important interactions involved in the binding mode of the ligand.   In general, the interaction point method for both FOMS and VAMS yielded the highest performance (cite table). The interaction point method is able to include specific areas where interactions are likely to occur between the receptor and ligand. Logically, this approach should yield the highest performance since it uses the most information involved in ligand-receptor interactions.     Fragment Selection:  The literature that cites the crystal structure from PDB should be conferred for fragment choice when determining the SMARTS expression for each protein target. Information about the important residues in the protein and which pocket(s)/interactions are the most important should be found. This information can easily come from the results section of the paper, especially in the sections highlighting the molecular interactions of the inhibitor (cite cathg), or in the section indicating the change in IC50 when certain residues are mutated (cite hivrt). Consider the most important interactions and have the SMARTS expression represent that fragment.     Cathg  Cathepsin G with its best approach and SMARTS expression, FOMS performed significantly better than random performance (cite stats table). For VAMS however, Cathepsin G with its best approach and SMARTS expression performed slightly worse than random chance (cite table). This indicates that for Cathepsin G, FOMS was a better performing screening method.  According to the study on the crystal structure of Cathepsin G with its inhibitor, Cathepsin G has 4 hydrophobic pockets that can be optimized for binding, each potentially being filled by a napthol group. Both FOMS and VAMS were pre-aligned to the deepest pocket in the active site, S1, the region that contributes the most to stabilizing the crystal structure (cite cathg). Although this is the most important interaction for a Cathepsin G inhibitor, filling the other pocket located at S3/S4 as referenced can contribute as much as 50 fold for increasing stability (cite cathg).   Docking of 1T32 with its active set from MUV using smina revealed 4 main pockets that were filled, the same shown in the previous study. The S1 pocket was filled the most frequently, with S3/S4 being filled second-most. This indicates that we pre-aligned our fragment to the most important pocket, however the S3/S4 pocket is also vital for Cathepsin G inhibition.   In general, less specific SMARTS expressions performed better for Cathepsin G for both FOMS and VAMS. (explanation for that).  ER-a  ER-a with its best approach and SMARTS expression performed significantly better than random performance for FOMS (cite stat table). For the VAMS method, the best approach and SMARTS expression for ER-a was not significantly better than random performance (cite stat table). This indicates that for ER-a, FOMS was a better performing screening method.   The most important interaction for an ER-a inhibitor is the 3-hydroxy interaction with Glu 353 and stacking effect of the phenyl group with Phe 404 (cite crystal, ER-a). Although these interactions were covered by our SMARTS expression and fragment pre-alignment, another important interaction occurring between a 17-hydroxy group and His 524 was not covered by either our SMARTS selection or fragment pre-alignment (cite crystal ER-a). Higher selectivity can also come from the addition of a group that performs hydrogen bonding with Leu 384. Because only two of the important interactions that ER-a inhibitors need to perform are covered by the SMARTS and pre-alignment, it makes sense that FOMS did not perform better than random chance (refer to Table I will make).   When docking with smina was performed on 2IOG and its set of active compounds from MUV, the two main interactions observed were with Glu353 and Arg394 by one hydrogen-bond donor/acceptor (visual inspection). This indicates that both residues are important to interact with for an ER-a inhibitor. Our fragment selection and pre-alignment covered this interaction. There were also interactions with Asp351 observed on the distal side with a hydrogen-bond donor, sometimes concurrent with the previous interactions. This indicates that this interaction is also key for the binding mode of ER-a inhibitors, and was not covered by our fragment pre-alignment. There were also 5 pockets that could potentially be filled. Our pre-alignment only covered the most occupied pocket among all active conformers docked.   More specific SMARTS expressions performed better for FOMS, but not VAMS for ER-a inhibitors (cite Table). For FOMS, the best fragment selection included pre-alignment to a phenolic SMARTS expression, which would align to the region that could form the interactions with Glu353 and Arg394. For this reason, it makes sense that this SMARTS expression and pre-alignment would produce the best performance for an ER-a inhibitor. (Explanation of why VAMS had better performance with lower specificity)    ER-a-p  ER-a with its best approach and SMARTS expression did not perform significantly better than random performance for the FOMS or VAMS method (cite table). Although FOMS was significantly better performing than VAMS, it still did not perform well enough for ER-a agonists to be considered an effective screening method.   Similar to ER-a, the interaction with Glu353 and Arg394 is vital for agonism of ER-a (cite crystal ER-a). There is also an indication that the stereochemistry is an important consideration for ER-a agonism when trying to avoid steric clash with W383, L536, and L539. The study reporting the crystal structure of the ER-a agonist cites, “…dihydrobenzoxathiin SERAMs [are] highly dependent on size and location of side chain substituents” (cite part 9). For this reason, it makes sense that our SMARTS expressions and fragment pre-alignments did not perform significantly better than random chance (cite table).   Docking of 1XP9 and its active set from MUV was performed using smina, and the two main interactions observed were with Glu353 and Arg394 concurrently (visual inspection). This indicates that both residues are important interactions for an ER-a agonist. In addition, there were 5 pockets available for filling. Three of these pockets were occupied by almost every conformer of the active compounds, with one of the main pockets being covered by fragment pre-alignment. The variation in which pockets needed to be filled indicates that there are a plethora of hydrophobic interactions that need to occur for ER-a agonism. This could explain the poor performance of both methods since it only pre-aligned to one of the many pockets.   For ER-a agonists, more specific SMARTS expressions had better performance for FOMS and VAMS (cite table). The best performing SMARTS expression covered the interaction with Glu353 and Arg394. This is what most likely gave it the best performance since interaction with these two residues are vital for ER-a agonism. (explanation for VAMS performance)  ER-b  ER-b with its best SMARTS expression and approach performed better than random performance for both FOMS and VAMS (cite stats table). VAMS performed slightly better than FOMS for ER-b, and the difference was significant (cite stats table). This indicates that VAMS is the best pre-screening method for ER-b.  As might be expected, the binding pocket of ER-b is very similar to ER-a, as is the mode of its inhibition. Similar to ER-a inhibitors, a hydrogen bond donor/acceptor interaction is necessary for interacting with Glu305 and Arg346. In addition, there is another hydrogen-bond donor interaction with His475 (cite ER-b). This indicates that both of these interactions are key to the binding mode of ER-b inhibitors. Only the Glu305 and Arg346 interactions are covered by our SMARTS expressions and fragment pre-alignments. This means that the other key interaction involving the conserved His on the distal side of the steroid would not be covered. This means that the FOMS approach could accept a compound with only the Glu and Arg interactions, while this type of compound would not actually bind (i. e. pick up a decoy). There are other important hydrophobic interactions evident from the binding of the inhibitor to ER-b, however the interactions mentioned above seem the most important (cite ER-b).   Docking of 1QKN was performed with its active compounds from MUV in smina. This revealed that there were 5 main pockets, with one being covered by the SMARTS expression and fragment pre-alignment. Because there are many pockets that could potentially be filled and weren’t also aligned to, it makes sense that the FOMS approach did not perform as well as VAMS.  The last issue is that rat ER-b was used for analysis. Although there is a high sequence homology, two residues in the ligand-binding cavity differ, which affect the hydrophobic interactions of active compounds. This could have also led to poor performance for both FOMS and VAMS.  More specific SMARTS expressions led to higher performance for both methods (cite table). This makes sense since the SMARTS expression and fragment pre-alignment covered the hydrogen-bonding interactions between Glu305 and Arg346, the most important interactions for ER-b inhibition.    FAK  FAK with its best SMARTS expression and approach performed significantly better than random performance for the FOMS method (cite stat table). Even with the best SMARTS expression and approach, the VAMS method performed worse than random performance for FAK (cite table).   FAK inhibition relies on hydrogen-bonding interactions with Cys502 and Arg426, along with an important hydrophobic interaction with Leu553 in the active site (cite FAK). Only the interaction with Cys502 is covered by the SMARTS expression and fragment pre-alignment. This indicates that other decoy compounds could be scored more favorably with FOMS since it would only focus on the fragment choice that highlights one of 3 key interactions. FAK inhibitors are supposed to be ATP-competitive, so it makes sense that only a small piece of the inhibitor cannot represent the entire binding mode of inhibition.   When docking with smina was performed on 3BZ3 and the active compounds that were covered by our SMARTS expressions, there were 5 pockets possible for filling observed, with only one covered by our fragment pre-alignment. In addition, hydrogen-bonding with Cys502 and Arg426 were observed frequently between active conformers. Our fragment pre-alignment and SMARTS expression would only cover the interaction with Cys502. Because there are more than one interaction and many pockets to be filled, it makes sense that FOMS was not able to yield high performance.   More specific SMARTS expressions led to higher performance for both methods (cite table). This makes sense as the best SMARTS expression and fragment pre-alignment covered the hydrogen-bonding interaction with Cys502, vital for FAK inhibition.     FXIa  The best SMARTS expression and approach for FXIa performed significantly better than random performance for the FOMS, but not the VAMS method (cite table). Inhibition of FXIa relies on hydrogen-bonding with many residues. Only one of the pockets and its interaction with a His57 by hydrogen bonding and a stacking interaction are covered by our SMARTS expression and fragment pre-alignment (cite FXIa). There are many other hydrogen-bond interactions that include hydrogen bonding with Asp189, Gly216, and Gly218 backbones that are not covered by our SMARTS expression and fragment pre-alignment. Similar to most other protein targets in the MUV dataset, FOMS would not fully capture the complete binding mode of compounds screened, which most likely led to its poor performance (cite table).   Docking of 2FDA with active compounds from the MUV dataset revealed 5 hydrophobic pockets, with only one being covered by our fragment pre-alignment. Many hydrogen-bonding interactions were revealed, including His57 and the inhibitor. Because our best SMARTS expression and fragment pre-alignment only covered the interaction with His57 and 1 of 5 potential pockets to be filled, it makes sense that FOMS did not yield high performance.   More specific SMARTS expressions led to higher performance for VAMS, whereas the least specific SMARTS expression led to highest performance for FOMS (cite table). (explanation for both?)  HIV-RT  The best SMARTS expression and approach for HIV-RT did not perform significantly better than random performance for both methods (cite table). Inhibition of HIV-RT relies on many key interactions, including a stacking interaction with Tyr188, hydrogen-bonding with Lys103, and many important hydrophobic interactions with Val and Phe residues (cite HIV). Single and double mutation varieites of HIV-RT were tested against known inhibitors, with the conserved Tyr residue mutation causing a 600 fold increase in IC50. Our original fragment pre-alignment and SMARTS expression only covered the conserved Lys interaction, however we then performed pre-alignment to the Tyr residue interaction and observed better performance (cite Table). As with most of the other targets in the MUV dataset, there are multiple interactions necessary for proper binding of an HIV-RT inhibitor. FOMS will only cover a small part of the necessary interaction.   Docking with 3DRP in smina for the active compounds from the MUV dataset were performed. This revealed 4 potential pockets to fill, with only one being covered by our fragment pre-alignment. In addition, hydrogen-bonding with Lys103 and potential stacking with Tyr188 were observed frequently. This indicates the importance of both interactions for proper HIV-RT binding. Since only one of the 4 pockets are covered by our fragment pre-alignment, it is possible that this is a contributing factor to the poor performance of FOMS.  Less specific SMARTS expressions led to higher performance for FOMS, whereas more specific SMARTS expressions led to higher performance for VAMS (cite table). (explanation for both?)    HSP 90  The best SMARTS expression and approach for Hsp90 performed significantly better than random performance for FOMS, but not for VAMS (cite table). Hsp 90 inhibitors interact with the ATP binding site. The most important interactions are with Asp93, Gly97, Lys58 , and Asn51 residues, with the study reporting the crystal structure highlighting Asp93 as the most important (cite HSP). Only the Asp93 interaction is covered by our SMARTS expressions and fragment pre-alignments, while the other interactions are also key for Hsp90 inhibition, with the Gly backbone being second in importance. As with other targets in the MUV dataset, the poor performance of FOMS relative to random performance can be partially attributed to the large number of interactions that are not covered by our fragment selection due to the nature of the ligand binding mode.  Docking with 2VCI in smina for the active compounds from the MUV dataset were performed. This revealed 4 potential hydrophobic pockets, with only one covered by our fragment pre-alignment. There were also many active conformers hydrogen-bonding with Asp93 and at least one other residue as well. These docking results indicate the importance of inclusion of every hydrogen-bonding interaction and more than one hydrophobic pocket for Hsp90 inhibition. Our SMARTS expression and fragment pre-alignment only covered one of these interactions and one of the pockets, which most likely led to FOMS not having high performance.   More specific SMARTS expressions led to higher performance for FOMS and VAMS (cite table). This makes sense as the more specific fragment choice would help both methods select compounds that can make the hydrogen-bond with Asp93.  PKA  The best SMARTS expression and approach for PKA performed significantly better than random performance for both FOMS and VAMS (cite stats table). The FOMS approach was significantly better than the VAMS approach for PKA (cite stats table).  Like many other targets in this dataset, PKA inhibitors interact with the ATP binding site. These inhibitors are relatively smaller in size compared to other inhibitors and have few electrostatic interactions. In the example of 1Q8U and its inhibitor, there is only one hydrogen bond with Val123, where the rest of the ligand makes multiple Van der Waals contacts with other parts of the target (cite PKA). SMARTS expressions and fragment pre-alignment covers the only electrostatic interaction and most of the hydrophobic interactions involved in PKA inhibitors, which most likely led to the high performance of FOMS (cite table).  More specific SMARTS expressions led to higher performance for FOMS (cite table). This makes sense as the more specific fragment choice would cover the only necessary hydrogen-bond interaction with the receptor. Less specific SMARTS expressions led to higher performance for the VAMS method (cite table). (explanation for why this is).  Rho-Kinase2  The best SMARTS expression and approach for Rho-Kinase2 performed significantly better than random performance for both FOMS and VAMS (cite stats table). The FOMS approach was significantly better than the VAMS approach for Rho-Kinase2 (cite stats table).   Similar to PKA, Rho-Kinase2 inhibitors are relatively small compared to other inhibtors and have few electrostatic interactions. In the case of 2H9V and its inhibitor, there are three hydrogen-bonding interactions with conserved Met, Asn, and Asp. There are also important hydrophobic interactions with other residues inside the pocket (cite Rho). Our SMARTS expression and fragment pre-alignment covers the hydrogen-bonding interaction with the Met and important hydrophobic interactions with conserved Val and Leu residues. This most likely led to the high performance of FOMS for Rho-Kinase2 (cite table).  More specific SMARTS expressions led to lower performance for both the FOMS and VAMS methods. (explanation for why this is ).