Unleashing the potential of digital pathology data by training computer-aided diagnosis models without human annotations.

The digitalization of clinical workflows and the increasing performance of deep learning algorithms are paving the way towards new methods for tackling cancer diagnosis. However, the availability of medical specialists to annotate digitized images and free-text diagnostic reports does not scale with the need for large datasets required to train robust computer-aided diagnosis methods that can target the high variability of clinical cases and data produced. This work proposes and evaluates an approach to eliminate the need for manual annotations to train computer-aided diagnosis tools in digital pathology. The approach includes two components, to automatically extract semantically meaningful concepts from diagnostic reports and use them as weak labels to train convolutional neural networks (CNNs) for histopathology diagnosis. The approach is trained (through 10-fold cross-validation) on 3'769 clinical images and reports, provided by two hospitals and tested on over 11'000 images from private and publicly available datasets. The CNN, trained with automatically generated labels, is compared with the same architecture trained with manual labels. Results show that combining text analysis and end-to-end deep neural networks allows building computer-aided diagnosis tools that reach solid performance (micro-accuracy = 0.908 at image-level) based only on existing clinical data without the need for manual annotations.

Selection and Characterization of Single Chain Antibody Fragments Specific for Hsp90 as a Potential Cancer Targeting Molecule.

Heat shock proteins play an essential role in facilitating malignant transformation and they have been recognized as important factors in human cancers. One of the key elements of the molecular chaperones machinery is Hsp90 and it has recently become a target for anticancer therapeutic approaches. The potential and importance of Hsp90-directed agents becomes apparent when one realizes that disruption of Hsp90 function may influence over 200 oncogenic client proteins. Here, we described the selection and characterization of Hsp90-specific antibody fragments from commercially available Tomlinson I and J phage display libraries. The affinities of Hsp90-binding scFv variants were measured using SPR method. Then, based on the best clone selected, we performed the affinity maturation procedure and obtained valuable Hsp90-specific clones. The selected binders were expressed and applied for immunostaining, ELISA and SPR analysis using model cancer cell lines. All performed experiments confirmed the ability of selected antibodies to interact with the Hsp90. Therefore, the presented Hsp90-specific scFv, might be a starting point for the development of a novel antibody-based strategy targeting cancer.

Design, expression and characterization of a highly stable tetratricopeptide-based protein scaffold for phage display application.

Tetratricopeptide repeat (TPR) is a structural motif mediating variety of protein-protein interactions. It has a high potential to serve as a small, stable and robust, non-immunoglobulin ligand binding scaffold. In this study, we showed the consensus approach to design the novel protein called designed tetratricopeptide repeat (dTPR), composed of three repeated 34 amino-acid tetratricopeptide motifs. The designed sequence was efficiently overexpressed in E. coli and purified to homogeneity. Recombinant dTPR is monomeric in solution and preserves its secondary structure within the pH range from 2.0 to 11.0. Its denaturation temperature at pH 7.5 is extremely high (104.5°C) as determined by differential scanning calorimetry. At extreme pH values the protein is still very stable: denaturation temperature is 90.1°C at pH 2.0 and 60.4°C at pH 11. Chemical unfolding of the dTPR is a cooperative, two-state process both at pH 7.5 and 2.0. The free energy of denaturation in the absence of denaturant equals to 15.0 kcal/mol and 13.5 kcal/mol at pH 7.5 and 2.0, respectively. Efficient expression and extraordinary biophysical properties make dTPR a promising framework for a biotechnological application, such as generation of specific ligand- binding molecules.

The role of novel genes rrp1(+) and rrp2(+) in the repair of DNA damage in Schizosaccharomyces pombe.

We identified two predicted proteins in Schizosaccharomyces pombe, Rrp1 (SPAC17A2.12) and Rrp2 (SPBC23E6.02) that share 34% and 36% similarity to Saccharomyces cerevisiae Ris1p, respectively. Ris1p is a DNA-dependent ATP-ase involved in gene silencing and DNA repair. Rrp1 and Rrp2 also share similarity with S. cerevisiae Rad5 and S. pombe Rad8, containing SNF2-N, RING finger and Helicase-C domains. To investigate the function of the Rrp proteins, we studied the DNA damage sensitivities and genetic interactions of null mutants with known DNA repair mutants. Single Deltarrp1 and Deltarrp2 mutants were not sensitive to CPT, 4NQO, CDPP, MMS, HU, UV or IR. The double mutants Deltarrp1 Deltarhp51 and Deltarrp2 Deltarhp51 plus the triple Deltarrp1 Deltarrp2 Deltarhp51 mutant did not display significant additional sensitivity. However, the double mutants Deltarrp1 Deltarhp57 and Deltarrp2 Deltarhp57 were significantly more sensitive to MMS, CPT, HU and IR than the Deltarhp57 single mutant. The checkpoint response in these strains was functional. In S. pombe, Rhp55/57 acts in parallel with a second mediator complex, Swi5/Sfr1, to facilitate Rhp51-dependent DNA repair. Deltarrp1 Deltasfr1 and Deltarrp2 Deltasfr1 double mutants did not show significant additional sensitivity, suggesting a function for Rrp proteins in the Swi5/Sfr1 pathway of DSB repair. Consistent with this, Deltarrp1 Deltarhp57 and Deltarrp2 Deltarhp57 mutants, but not Deltarrp1 Deltasfr1 or Deltarrp2 Deltasfr1 double mutants, exhibited slow growth and aberrations in cell and nuclear morphology that are typical of Deltarhp51.