<HashMap><database>biostudies-literature</database><scores/><additional><submitter>Liu Z</submitter><funding>Natural Science Foundation of Shanghai</funding><funding>National Natural Science Foundation of China</funding><funding>National Key R &amp;amp;amp; D Program of China</funding><funding>Shanghai Sailing Program</funding><pagination>1454</pagination><full_dataset_link>https://www.ebi.ac.uk/biostudies/studies/S-EPMC9598995</full_dataset_link><repository>biostudies-literature</repository><omics_type>Unknown</omics_type><volume>11(10)</volume><pubmed_abstract>Though AlphaFold2 has attained considerably high precision on protein structure prediction, it is reported that directly inputting coordinates into deep learning networks cannot achieve desirable results on downstream tasks. Thus, how to process and encode the predicted results into effective forms that deep learning models can understand to improve the performance of downstream tasks is worth exploring. In this study, we tested the effects of five processing strategies of coordinates on two single-sequence protein binding site prediction tasks. These five strategies are spatial filtering, the singular value decomposition of a distance map, calculating the secondary structure feature, and the relative accessible surface area feature of proteins. The computational experiment results showed </pubmed_abstract><journal>Biology</journal><pubmed_title>Evaluation of the Effectiveness of Derived Features of AlphaFold2 on Single-Sequence Protein Binding Site Prediction.</pubmed_title><pmcid>PMC9598995</pmcid><funding_grant_id>nos. 81971292, 82150610506, 12104088</funding_grant_id><funding_grant_id>81971292</funding_grant_id><funding_grant_id>No: 21ZR1428600</funding_grant_id><funding_grant_id>12104088</funding_grant_id><funding_grant_id>21ZR1428600</funding_grant_id><funding_grant_id>82150610506</funding_grant_id><funding_grant_id>2022ZD0209100</funding_grant_id><funding_grant_id>No: 19YF1400200</funding_grant_id><funding_grant_id>19YF1400200</funding_grant_id><pubmed_authors>Li W</pubmed_authors><pubmed_authors>Pan W</pubmed_authors><pubmed_authors>Liu Z</pubmed_authors><pubmed_authors>Zhen X</pubmed_authors><pubmed_authors>Liang J</pubmed_authors><pubmed_authors>Yuan K</pubmed_authors><pubmed_authors>Xu F</pubmed_authors><pubmed_authors>Lin GN</pubmed_authors><pubmed_authors>Cai W</pubmed_authors></additional><is_claimable>false</is_claimable><name>Evaluation of the Effectiveness of Derived Features of AlphaFold2 on Single-Sequence Protein Binding Site Prediction.</name><description>Though AlphaFold2 has attained considerably high precision on protein structure prediction, it is reported that directly inputting coordinates into deep learning networks cannot achieve desirable results on downstream tasks. Thus, how to process and encode the predicted results into effective forms that deep learning models can understand to improve the performance of downstream tasks is worth exploring. In this study, we tested the effects of five processing strategies of coordinates on two single-sequence protein binding site prediction tasks. These five strategies are spatial filtering, the singular value decomposition of a distance map, calculating the secondary structure feature, and the relative accessible surface area feature of proteins. The computational experiment results showed </description><dates><release>2022-01-01T00:00:00Z</release><publication>2022 Oct</publication><modification>2026-06-21T03:08:49.775Z</modification><creation>2024-11-19T16:30:07.826Z</creation></dates><accession>S-EPMC9598995</accession><cross_references><pubmed>36290358</pubmed><doi>10.3390/biology11101454</doi></cross_references></HashMap>