Protein modeling quality

Protein modeling quality is an important part of protein structure prediction. The new quality predictor ProQ3D was developed using deep learning techniques and could improve prediction accuracy relative to previous methods (Uziela et al 2018). ProQ3D was shown to perform at top in the CASP13 evaluation.

We contributed software and methods to assess the inter-rater agreement on prostate biopsies using the ImageBase database. We showed that there is poor agreement between pathologists on the Gleason grading of prostate biopsies (Egevad et al 2018). As a result of these deficiencies, we have been developing methods to use convolutional deep learning networks to analyse prostate biopsy images. Publications from the deep learning are due in 2019.

Two SeRC PhD students with Keith Humphreys, Gabriel Isheden and Rickard Strandberg, gained student conference prizes for best papers on their modelling of tumour growth.

  • Egevad L, Delahunt B, Berney DM, Bostwick DG, Cheville J, Comperat E, Evans AJ, Fine SW, Grignon DJ, Humphrey PA, Hörnblad J, Iczkowski KA, Kench JG, Kristiansen G, Leite KRM, Magi-Galluzzi C, McKenney JK, Oxley J, Pan CC, Samaratunga H, Srigley JR, Takahashi H, True LD, Tsuzuki T, van der Kwast T, Varma M, Zhou M, Clements M. Utility of Pathology Imagebase for standardisation of prostate cancer grading. Histopathology. 2018 Jul;73(1):8-18. doi: 10.1111/his.13471. Epub 2018 Mar 5. PubMed PMID: 29359484.
  • Uziela, K., Menendez Hurtado, D., Shu, N., Wallner, B. and Elofsson, A. (2018) Improved protein model quality assessments by changing the target function. Proteins 86 (6) : 654-663.