(B-59) Script for Semi-automatic Analysis of SAXS Data for Polymer-Protein Hybrids

Introduction:: Protein-polymer hybrids (PPHs) can be used to stabilize proteins in otherwise denaturing environments. PPHs are synthesized by the Gormley group using a design-test-build approach driven by machine learning (ML) and robotics (e.g., liquid handling robot) (Figure 1) [1]. Small angle X-ray scattering (SAXS) is a characterization technique to probe the structure of macromolecules. SAXS aids the design of PPHs by probing polymer integration to protein surface through parameters such as radius of gyration (Rg) and pair distribution function (PDF) analysis. Common software used for SAXS data analysis include BioXTAS and SasView. Nonetheless, there is a need to adapt these to better suit the current high-throughput PPH synthesis pipeline by reducing the need for user inputs/inspection of data and facilitating database generation for subsequent training of ML models in compatible formats.


Materials and Methods:: Open-source software BioXTAS Raw (Raw) was used to create the script for semi-automatic SAXS data analysis. In addition to this, Guinier peak analysis (GPA) is used for rapid inspection of data quality as this can suggest data quality issues such as parameter errors, or issues with low resolution data [2]. The python script was created using google colab. The script was tested using 2,009 experimental SAXS profiles from the Small Angle Scattering Biological Data Bank (SASBDB)[3], as these have high resolution the profiles were sampled to have lower resolution with q­_min = 0.0005.

Results, Conclusions, and Discussions:: Results and Discussion: The script was tested with the SASBDB profiles and compared to a single run of BioXTAS Raw where the auto Guinier function was run first then followed by PDF using the BIFT function. The results from this initial large-scale test showed that out of the 2,009 experimental profiles, 1,759 files were analyzed as expected (88%), with profiles with data (e.g., no Guinier region) being rejected. The other file names are recorded to be analyzed by a user as these require close inspection. Furthermore, the Rg determined by the script was within 12% of the values reported to the SASBDB (Figure 2) with 19 files having an absolute difference greater than 15%. The output from the script can be converted into a PDF or HTML file for user inspection of results and the data is summarized into a pandas’ data frame which can be converted to an excel sheet.


Conclusion: PPHs are of importance to deliver stabilized proteins that withstand otherwise denaturing environments. To accomplish this, the current high-throughput synthesis pipeline needs to be coupled with a high-throughput data analysis pipeline that can provide structural parameters from SAXS data to further train the model and thus aid in generating new improved PPHs through this design-test-build process. Whereas this semi-automatic approach can aid in the high-throughput generation of data for ML model training.


Acknowledgements (Optional): : The authors acknowledge NIH NIGMS R35GM138296

References (Optional): : [1] Tamasi, M., Adv. Mater., 2022 [2] Putnam, C., JAC., 2016 [3] Kikhney, A.,Protein Science., 2019