Bart van Oort

Machine Learning (ML) projects incur novel challenges in their development and productionisation over traditional software applications, though established principles and best practices in ensuring the project's software quality still apply. While using static analysis to catch code smells has been shown to improve software quality attributes, it is only a small piece of the software quality puzzle, especially in the case of ML projects given their additional challenges and lower degree of Software Engineering (SE) experience in the data scientists that develop them. We introduce the novel concept of project smells which consider deficits in project management as a more holistic perspective on software quality in ML projects. An open-source static analysis tool mllint was also implemented to help detect and mitigate these. Our research evaluates this novel concept of project smells in the industrial context of ING, a global bank and large software- and data-intensive organisation. We also investigate the perceived importance of these project smells for proof-of-concept versus production-ready ML projects, as well as the perceived obstructions and benefits to using static analysis tools such as mllint. Our findings indicate a need for context-aware static analysis tools, that fit the needs of the project at its current stage of development, while requiring minimal configuration effort from the user.

Photobiocatalysis holds great promise towards the development of sustainable and environmentally friendly processes, harnessing light to drive biocatalytic reactions. However, photobiocatalysis at the interface of insoluble substrates, such as cellulose has not been studied in much detail. In this context, the catalytic enhancement of lytic polysaccharide monooxygenases (LPMOs) by light is of great interest to the biorefinery field due to their capacity to oxidatively cleave such recalcitrant polysaccharides which can facilitate the degradation of lignocellulose. It has previously been reported that light-driven LPMO-reactions have a huge catalytic potential but effective continuous illumination in reactors may be challenging. Therefore, we investigated the impact of intermittent illumination. We show that illumination intervals as short as 1 s/min enables LPMO-catalysis on phosphoric acid swollen cellulose (PASC) to the same level as continuous illumination. Additionally, time-resolved measurements indicate that reductant depletion, and not enzyme inactivation, limits light-driven LPMO reactions. This study shows that a 60-fold reduction in illumination time enhances LPMO-catalysis while protecting reaction elements, e.g. the reductant. Most importantly, the significant enhancement of LPMO-catalysis with minimal and intermittent illumination is promising towards an application of photobiocatalytic depolymerization of lignocellulose where shading and light-scattering minimize light availability and continuity.

Lytic polysaccharide monooxygenases (LPMOs) are copper-containing enzymes which promote the degradation of recalcitrant polysaccharides like cellulose or chitin. Here, we have investigated the thermostability of an LPMO from Thermoascus aurantiacus (TaLPMO9A). TaLPMO9A was found to retain most of its initial activity after incubating at 100 °C while its apparent melting temperature (T_m) is 69 °C at neutral pH. Interestingly, our studies show that holoTaLPMO9A, apoTaLPMO9A and deglycosylated TaLPMO9A can fold back to their original conformation upon lowering the temperature. In the presence of β-mercaptoethanol the protein does not refold. Activity of TaLPMO9A and refolded TaLPMO9A was studied by an Amplex® Red assay as well as by TaLPMO9A catalysed oxidation of phosphoric acid swollen cellulose (PASC). These studies confirm the functional regain of TaLPMO9A activity upon going through one cycle of unfolding and refolding. The thermal unfolding and refolding of TaLPMO9A was measured spectroscopically. Utilizing the two-state model, detailed thermodynamic parameters were obtained for holoTaLPMO. Furthermore, we have investigated the kinetics of TaLPMO9A unfolding and refolding. Our results have implications in understanding LPMO stability, which is crucial for the efficient application of LPMOs as biocatalysts during biomass degradation.

Cyanobacteria have the potential to become next-generation cell factories due to their ability to use CO2, light and inorganic nutrients to produce a range of biomolecules of commercial interest. Synechococcus elongatus UTEX 2973, in particular, is a fast-growing, genetically tractable, cyanobacterium that has garnered attention as a potential biotechnological chassis. To establish this unique strain as a host for heterologous protein production, we aimed to demonstrate expression and secretion of the industrially relevant TfAA10A, a lytic polysaccharide monooxygenase from the Gram-positive bacterium Thermobifida fusca. Two variations of TfAA10A were successfully expressed in S. elongatus UTEX 2973: One containing the native N-terminal, Sec-targeted, signal peptide and a second with a Tat-targeted signal peptide from the Escherichia coli trimethylamine-N-oxide reductase (TorA). Although the TorA signal peptide correctly targeted the protein to the plasma membrane, the majority of the TorA-TfAA10A was found unprocessed in the plasma membrane with a small fraction of the mature protein ultimately translocated to the periplasm. The native Sec signal peptide allowed for efficient secretion of TfAA10A into the medium with virtually no protein being found in the cytosol, plasma membrane or periplasm. TfAA10A was demonstrated to be correctly cleaved and active on the model substrate phosphoric acid swollen cellulose. Additionally, expression and secretion only had a minor impact on cell growth. The secretion yield was estimated at 779 ± 40 µg L−1 based on densitometric analysis. To our knowledge, this is the highest secretion yield ever registered in cyanobacteria. We have shown for the first time high-titer expression and secretion of an industrially relevant and catalytically active enzyme in S. elongatus UTEX 2973. This proof-of-concept study will be valuable for the development of novel and sustainable applications in the fields of bioremediation and biocatalysis.

In contrast to single cellular species, detailed information is lacking on the processes of photosynthetic acclimation for colonial algae, although these algae are important for biofuel production, ecosystem biodiversity, and wastewater treatment. To investigate differences between single cellular and colonial species, we studied the regulation of photosynthesis and photoprotection during photoacclimation for the colonial green alga Botryococcus braunii and made a comparison with the properties of the single cellular species Chlamydomonas reinhardtii. We show that B. braunii shares some high-light (HL) photoacclimation strategies with C. reinhardtii and other frequently studied green algae: decreased chlorophyll content, increased free carotenoid content, and increased nonphotochemical quenching (NPQ). Additionally, B. braunii has unique HL photoacclimation strategies, related to its colonial form: strong internal shading by an increase of the colony size and the accumulation of extracellular echinenone (a ketocarotenoid). HL colonies are larger and more spatially heterogenous than low-light colonies. Compared with surface cells, cells deeper inside the colony have increased pigmentation and larger photosystem II antenna size. The core of the largest of the HL colonies does not contain living cells. In contrast with C. reinhardtii, but similar to other biofilm-forming algae, NPQ capacity is substantial in low light. In HL, NPQ amplitude increases, but kinetics are unchanged. We discuss possible causes of the different acclimation responses of C. reinhardtii and B. braunii. Knowledge of the specific photoacclimation processes for this colonial green alga further extends the view of the diversity of photoacclimation strategies in photosynthetic organisms.

Photoprotective non-photochemical quenching (NPQ) of absorbed solar energy is vital for survival of photosynthetic organisms, and NPQ modifications significantly improve plant productivity. However, the exact NPQ quenching mechanism is obscured by discrepancies between reported mechanisms, involving xanthophyll-chlorophyll (Xan-Chl) and Chl-Chl interactions. We present evidence of an experimental artefact that may explain the discrepancies: strong laser pulses lead to the formation of a novel electronic species in the major plant light-harvesting complex (LHCII). This species evolves from a high excited state of Chl a and is absent with weak laser pulses. It resembles an excitonically coupled heterodimer of Chl a and lutein (or other Xans at site L1), and acts as de-excitation channel. Laser powers, and consequently amounts of artefact, vary strongly between NPQ studies, thereby explaining contradicting spectral signatures attributed to NPQ. Our results offer pathways towards unveiling NPQ mechanisms, and highlight the necessity of careful attention to laser-induced artefacts.

To avoid photodamage plants regulate the amount of excitation energy in the membrane at the level of the light-harvesting complexes (LHCs). It has been proposed that the energy absorbed in excess is dissipated via protein conformational changes of individual LHCs. However, the exact quenching mechanism remains unclear. Here we study the mechanism of quenching in LHCs that bind a single carotenoid species and are constitutively in a dissipative conformation. Via femtosecond spectroscopy we resolve a number of carotenoid dark states, demonstrating that the carotenoid is bound to the complex in different conformations. Some of those states act as excitation energy donors for the chlorophylls, whereas others act as quenchers. Via in silico analysis we show that structural changes of carotenoids are expected in the LHC protein domains exposed to the chloroplast lumen, where acidification triggers photoprotection in vivo. We propose that structural changes of LHCs control the conformation of the carotenoids, thus permitting access to different dark states responsible for either light harvesting or photoprotection.

The colonial green alga Botryococcus braunii (BB) is a potential source of biofuel due to its natural high hydrocarbon content. Unfortunately, its slow growth limits its biotechnological potential. Understanding its photosynthetic machinery could help to identify possible growth limitations. Here, we present the first study on BB light-harvesting complexes (LHCs). We purified two LHC fractions containing the complexes in monomeric and trimeric form. Both fractions contained at least two proteins with molecular weight (MW) around 25 kDa. The chlorophyll composition is similar to that of the LHCII of plants; in contrast, the main xanthophyll is loroxanthin, which substitutes lutein in most binding sites. Circular dichroism and 77 K absorption spectra lack typical differences between monomeric and trimeric complexes, suggesting that intermonomer interactions do not play a role in BB LHCs. This is in agreement with the low stability of the BB LHCII trimers as compared to the complexes of plants, which could be related to loroxanthin binding in the central (L1 and L2) binding sites. The properties of BB LHCII are similar to those of plant LHCII, indicating a similar pigment organization. Differences are a higher content of red chlorophyll a, similar to plant Lhcb3. These differences and the different Xan composition had no effect on excitation energy transfer or fluorescence lifetimes, which were similar to plant LHCII.