FEP Augmentation as a Means to Solve Data Paucity Problems for Machine Learning in Chemical Biology.

In the realm of medicinal chemistry, the primary objective is to swiftly optimize a multitude of chemical properties of a set of compounds to yield a clinical candidate poised for clinical trials. In recent years, two computational techniques, machine learning (ML) and physics-based methods, have evolved substantially and are now frequently incorporated into the medicinal chemist's toolbox to enhance the efficiency of both hit optimization and candidate design. Both computational methods come with their own set of limitations, and they are often used independently of each other. ML's capability to screen extensive compound libraries expediently is tempered by its reliance on quality data, which can be scarce especially during early-stage optimization. Contrarily, physics-based approaches like free energy perturbation (FEP) are frequently constrained by low throughput and high cost by comparison; however, physics-based methods are capable of making highly accurate binding affinity predictions. In this study, we harnessed the strength of FEP to overcome data paucity in ML by generating virtual activity data sets which then inform the training of algorithms. Here, we show that ML algorithms trained with an FEP-augmented data set could achieve comparable predictive accuracy to data sets trained on experimental data from biological assays. Throughout the paper, we emphasize key mechanistic considerations that must be taken into account when aiming to augment data sets and lay the groundwork for successful implementation. Ultimately, the study advocates for the synergy of physics-based methods and ML to expedite the lead optimization process. We believe that the physics-based augmentation of ML will significantly benefit drug discovery, as these techniques continue to evolve.

Shape, size, and quantity of ingested external abrasives influence dental microwear texture formation in guinea pigs.

Food processing wears down teeth, thus affecting tooth functionality and evolutionary success. Other than intrinsic silica phytoliths, extrinsic mineral dust/grit adhering to plants causes tooth wear in mammalian herbivores. Dental microwear texture analysis (DMTA) is widely applied to infer diet from microscopic dental wear traces. The relationship between external abrasives and dental microwear texture (DMT) formation remains elusive. Feeding experiments with sheep have shown negligible effects of dust-laden grass and browse, suggesting that intrinsic properties of plants are more important. Here, we explore the effect of clay- to sand-sized mineral abrasives (quartz, volcanic ash, loess, kaolin) on DMT in a controlled feeding experiment with guinea pigs. By adding 1, 4, 5, or 8% mineral abrasives to a pelleted base diet, we test for the effect of particle size, shape, and amount on DMT. Wear by fine-grained quartz (>5/<50 µm), loess, and kaolin is not significantly different from the abrasive-free control diet. Fine silt-sized quartz (∼5 µm) results in higher surface anisotropy and lower roughness (polishing effect). Coarse-grained volcanic ash leads to significantly higher complexity, while fine sands (130 to 166 µm) result in significantly higher roughness. Complexity and roughness values exceed those from feeding experiments with guinea pigs who received plants with different phytolith content. Our results highlight that large (>95-µm) external silicate abrasives lead to distinct microscopic wear with higher roughness and complexity than caused by mineral abrasive-free herbivorous diets. Hence, high loads of mineral dust and grit in natural diets might be identified by DMTA, also in the fossil record.

Forage silica and water content control dental surface texture in guinea pigs and provide implications for dietary reconstruction.

Recent studies have shown that phytoliths are softer than dental enamel but still act as abrasive agents. Thus, phytolith content should be reflected in dental wear. Because native phytoliths show lower indentation hardness than phytoliths extracted by dry ashing, we propose that the hydration state of plant tissue will also affect dental abrasion. To assess this, we performed a controlled feeding experiment with 36 adult guinea pigs, fed exclusively with three different natural forages: lucerne, timothy grass, and bamboo with distinct phytolith/silica contents (lucerne < grass < bamboo). Each forage was fed in fresh or dried state for 3 weeks. We then performed 3D surface texture analysis (3DST) on the upper fourth premolar. Generally, enamel surface roughness increased with higher forage phytolith/silica content. Additionally, fresh and dry grass feeders displayed differences in wear patterns, with those of fresh grass feeders being similar to fresh and dry lucerne (phytolith-poor) feeders, supporting previous reports that "fresh grass grazers" show less abrasion than unspecialized grazers. Our results demonstrate that not only phytolith content but also properties such as water content can significantly affect plant abrasiveness, even to such an extent that wear patterns characteristic for dietary traits (browser-grazer differences) become indistinguishable.

Movement analysis of primate molar teeth under load using synchrotron X-ray microtomography.

Mammalian teeth have to sustain repetitive and high chewing loads without failure. Key to this capability is the periodontal ligament (PDL), a connective tissue containing a collagenous fibre network which connects the tooth roots to the alveolar bone socket and which allows the teeth to move when loaded. It has been suggested that rodent molars under load experience a screw-like downward motion but it remains unclear whether this movement also occurs in primates. Here we use synchroton micro-computed tomography paired with an axial loading setup to investigate the form-function relationship between tooth movement and the morphology of the PDL space in a non-human primate, the mouse lemur (Microcebus murinus). The loading behavior of both mandibular and maxillary molars showed a three-dimensional movement with translational and rotational components, which pushes the tooth into the alveolar socket. Moreover, we found a non-uniform PDL thickness distribution and a gradual increase in volumetric proportion of the periodontal vasculature from cervical to apical. Our results suggest that the PDL morphology may optimize the three-dimensional tooth movement to avoid high stresses under loading.

Dental microwear texture gradients in guinea pigs reveal that material properties of the diet affect chewing behaviour.

Dental microwear texture analysis (DMTA) is widely used for diet inferences in extant and extinct vertebrates. Often, a reference tooth position is analysed in extant specimens, while isolated teeth are lumped together in fossil datasets. It is therefore important to test whether dental microwear texture (DMT) is tooth position specific and, if so, what causes the differences in wear. Here, we present results from controlled feeding experiments with 72 guinea pigs, which received either fresh or dried natural plant diets of different phytolith content (lucerne, grass, bamboo) or pelleted diets with and without mineral abrasives (frequently encountered by herbivorous mammals in natural habitats). We tested for gradients in dental microwear texture along the upper cheek tooth row. Regardless of abrasive content, guinea pigs on pelleted diets displayed an increase in surface roughness along the tooth row, indicating that posterior tooth positions experience more wear compared with anterior teeth. Guinea pigs feedings on plants of low phytolith content and low abrasiveness (fresh and dry lucerne, fresh grass) showed almost no DMT differences between tooth positions, while individuals feeding on more abrasive plants (dry grass, fresh and dry bamboo) showed a gradient of decreasing surface roughness along the tooth row. We suggest that plant feeding involves continuous intake and comminution by grinding, resulting in posterior tooth positions mainly processing food already partly comminuted and moistened. Pelleted diets require crushing, which exerts higher loads, especially on posterior tooth positions, where bite forces are highest. These differences in chewing behaviour result in opposing wear gradients for plant versus pelleted diets.

Dust and grit matter: abrasives of different size lead to opposing dental microwear textures in experimentally fed sheep (Ovis aries).

External abrasives ingested along with the herbivore diet are considered main contributors to dental wear, though how the different sizes and concentrations of these abrasives influence wear remains unclear. Dental microwear texture analysis (DMTA) is an established method for dietary reconstruction which describes a tooth's surface topography on a micrometre scale. The method has yielded conflicting results as to the effect of external abrasives. In the present study, a feeding experiment was performed on sheep (Ovis aries) fed seven diets of different abrasiveness. Our aim was to discern the individual effects of size (4, 50 and 130 µm) and concentration (0%, 4% and 8% of dry matter) of abrasives on dental wear, applying DMTA to four tooth positions. Microwear textures differed between individual teeth, but surprisingly, showed no gradient along the molar tooth row, and the strongest differentiation of experimental groups was achieved when combining data of all maxillary molars. Overall, a pattern of increasing height, volume and complexity of the tooth's microscopic surface appeared with increasing size of dietary abrasives, and when compared with the control, the small abrasive diets showed a polishing effect. The results indicate that the size of dietary abrasives is more important for dental microwear texture traces than their concentration, and that different sizes can have opposing effects on the dietary signal. The latter finding possibly explains conflicting evidence from previous experimental DMTA applications. Further exploration is required to understand whether and how microscopic traces created by abrasives translate quantitatively to tissue loss.

Synthesis and Antiviral Activity of a Series of 2'-C-Methyl-4'-thionucleoside Monophosphate Prodrugs.

The NS5B RNA-dependent RNA polymerase of the hepatitis C virus (HCV) is a validated target for nucleoside antiviral drug therapy. We endeavored to synthesize and test a series of 4'-thionucleosides with a monophosphate prodrug moiety for their antiviral activity against HCV and other related viruses in the Flaviviridae family. Nucleoside analogs were prepared via the stereoselective Vorbrüggen glycosylation of various nucleobases with per-acetylated 2-C-methyl-4-thio-d-ribose built in a 10-step synthetic sequence from the corresponding ribonolactone. Conjugation of the thionucleoside to a ProTide phosphoramidate allowed for evaluation of the prodrugs in the cellular HCV replicon assay with anti-HCV activities ranging from single-digit micromolar (µM) to >200 µM. The diminished anti-HCV potency of our best compound compared to its 4'-oxo congener is the subject of ongoing research in our lab and is proposed to stem from changes in sugar geometry imparted by the larger sulfur atom.

Dental microwear texture reflects dietary tendencies in extant Lepidosauria despite their limited use of oral food processing.

Lepidosauria show a large diversity in dietary adaptations, both among extant and extinct tetrapods. Unlike mammals, Lepidosauria do not engage in sophisticated mastication of their food and most species have continuous tooth replacement, further reducing the wear of individual teeth. However, dietary tendency estimation of extinct lepidosaurs usually rely on tooth shape and body size, which allows only for broad distinction between faunivores and herbivores. Microscopic wear features on teeth have long been successfully applied to reconstruct the diet of mammals and allow for subtle discrimination of feeding strategies and food abrasiveness. Here, we present, to our knowledge, the first detailed analysis of dental microwear texture on extant lepidosaurs using a combination of 46 surface texture parameters to establish a framework for dietary tendency estimation of fossil reptilian taxa. We measured dental surface textures of 77 specimens, belonging to herbivorous, algaevorous, frugivorous, carnivorous, ovivorous, insectivorous, molluscivorous, as well as omnivorous species. Carnivores show low density and shallow depth of furrows, whereas frugivores are characterized by the highest density of furrows. Molluscivores show the deepest wear features and highest roughness, herbivores have lower surface roughness and shallower furrows compared to insectivores and omnivores, which overlap in all parameters. Our study shows that despite short food-tooth interaction, dental surface texture parameters enable discrimination of several feeding strategies in lepidosaurs. This result opens new research avenues to assess diet in a broad variety of extant and extinct non-mammalian taxa including dinosaurs and early synapsids.

Error Tolerance of Machine Learning Algorithms across Contemporary Biological Targets.

Machine learning continues to make strident advances in the prediction of desired properties concerning drug development. Problematically, the efficacy of machine learning in these arenas is reliant upon highly accurate and abundant data. These two limitations, high accuracy and abundance, are often taken together; however, insight into the dataset accuracy limitation of contemporary machine learning algorithms may yield insight into whether non-bench experimental sources of data may be used to generate useful machine learning models where there is a paucity of experimental data. We took highly accurate data across six kinase types, one GPCR, one polymerase, a human protease, and HIV protease, and intentionally introduced error at varying population proportions in the datasets for each target. With the generated error in the data, we explored how the retrospective accuracy of a Naïve Bayes Network, a Random Forest Model, and a Probabilistic Neural Network model decayed as a function of error. Additionally, we explored the ability of a training dataset with an error profile resembling that produced by the Free Energy Perturbation method (FEP+) to generate machine learning models with useful retrospective capabilities. The categorical error tolerance was quite high for a Naïve Bayes Network algorithm averaging 39% error in the training set required to lose predictivity on the test set. Additionally, a Random Forest tolerated a significant degree of categorical error introduced into the training set with an average error of 29% required to lose predictivity. However, we found the Probabilistic Neural Network algorithm did not tolerate as much categorical error requiring an average of 20% error to lose predictivity. Finally, we found that a Naïve Bayes Network and a Random Forest could both use datasets with an error profile resembling that of FEP+. This work demonstrates that computational methods of known error distribution like FEP+ may be useful in generating machine learning models not based on extensive and expensive in vitro-generated datasets.

The Bioactive Protein-Ligand Conformation of GluN2C-Selective Positive Allosteric Modulators Bound to the NMDA Receptor.

N-methyl-d-aspartate (NMDA) receptors are ligand-gated, cation-selective channels that mediate a slow component of excitatory synaptic transmission. Subunit-selective positive allosteric modulators of NMDA receptor function have therapeutically relevant effects on multiple processes in the brain. A series of pyrrolidinones, such as PYD-106, that selectively potentiate NMDA receptors that contain the GluN2C subunit have structural determinants of activity that reside between the GluN2C amino terminal domain and the GluN2C agonist binding domain, suggesting a unique site of action. Here we use molecular biology and homology modeling to identify residues that line a candidate binding pocket for GluN2C-selective pyrrolidinones. We also show that occupancy of only one site in diheteromeric receptors is required for potentiation. Both GluN2A and GluN2B can dominate the sensitivity of triheteromeric receptors to eliminate the actions of pyrrolidinones, thus rendering this series uniquely sensitive to subunit stoichiometry. We experimentally identified NMR-derived conformers in solution, which combined with molecular modeling allows the prediction of the bioactive binding pose for this series of GluN2C-selective positive allosteric modulators of NMDA receptors. These data advance our understanding of the site and nature of the ligand-protein interaction for GluN2C-selective positive allosteric modulators for NMDA receptors.

Controlled feeding experiments with diets of different abrasiveness reveal slow development of mesowear signal in goats (Capra aegagrus hircus).

Dental mesowear is applied as a proxy to determine the general diet of mammalian herbivores based on tooth-cusp shape and occlusal relief. Low, blunt cusps are considered typical of grazers and high, sharp cusps typical of browsers. However, how internal or external abrasives impact mesowear, and the time frame the wear signature takes to develop, still need to be explored. Four different pelleted diets of increasing abrasiveness (lucerne, grass, grass and rice husks, and grass, rice husks and sand) were fed to four groups of a total of 28 adult goats in a controlled feeding experiment over a 6-month period. Tooth morphology was captured by medical CT scans at the beginning and end of the experiment. These scans, as well as the crania obtained post mortem, were scored using the mesowear method. Comparisons between diet groups showed few significant differences after 6 months, irrespective of whether CT scans or the real teeth were scored. Only when assessing the difference in signal between the beginning and the end of the experiment did relevant, significant diet-specific effects emerge. Diets containing lower phytolith content caused a more pronounced change in mesowear towards sharper cusps/higher reliefs, while the feed containing sand did not result in more extreme changes in mesowear when compared with the same feed without sand. Our experiment suggests that the formation of a stable and hence reliable mesowear signal requires more time to develop than 6 months.

A Machine Learning Approach for Predicting HIV Reverse Transcriptase Mutation Susceptibility of Biologically Active Compounds.

HIV resistance emerging against antiretroviral drugs represents a great threat to the continued prolongation of the lifespans of HIV-infected patients. Therefore, methods capable of predicting resistance susceptibility in the development of compounds are in great need. By targeting the major reverse transcription residues Y181, K103, and L100, we used the biological activities of compounds against these enzymes and the wild-type reverse transcriptase to create Naïve Bayes Networks. Through this machine learning approach, we could predict, with high accuracy, whether a compound would be susceptible to a loss of potency due to resistance. Also, we could perfectly predict retrospectively whether compounds would be susceptible to both a K103 mutant RT and a Y181 mutant RT. In the study presented here, our method outperformed a traditional molecular mechanics approach. This method should be of broad interest beyond drug discovery efforts, and serves to expand the utility of machine learning for the prediction of physical, chemical, or biological properties using the vast information available in the literature.

Analyzing the sclerocarpy adaptations of the Pitheciidae mandible.

Primates are interpreted to be ancestrally adapted to frugivory, although some modern groups show clear adaptations to other diets. Among them, pitheciids stand out for specifically predating seeds. This dietary specialization is known as sclerocarpy and refers to the extraction of seeds from surrounding hard tissues using the anterior dentition followed by the mastication of seeds by the molars. It has been proposed that Callicebus-Pithecia-Chiropotes-Cacajao represent a morphocline of increasingly specialized anatomical traits for sclerocarpic foraging. This study addresses whether there is a sclerocarpic specialization gradient in the mandibular morphology of pitheciids. Finite element analysis (FEA) was used to simulate two biting scenarios and the obtained stress values were compared between different pitheciids. Geometric morphometrics (GM) were used to display the morphological variation of this group. No support was found for the morphocline hypothesis from a biomechanical viewpoint since all pitheciins showed similar stress values and on average Chiropotes rather than Cacajao exhibited the strongest mandible. From a morphological perspective, it was found that there is indeed relative "robusticity" continuum in the pitheciid mandible for some aspects of shape as expected for the morphocline hypothesis, but this gradient could be related to other factors rather than sclerocarpic specialization. The present results are expected to contribute to a better insight regarding the ecomorphological relationship between mandibular morphology and mechanical performance among pitheciids.

Regiochemistry discoveries in the use of isoxazole as a handle for the rapid construction of an all-carbon macrocyclic precursor in the synthetic studies of celastrol.

We have developed a convergent synthetic route to an all-carbon, 14-membered Z,E-macrocyclic bis-enone during our synthetic study of celastrol. The 1,3-dipolar cycloaddition of nitrile oxide and alkyne was employed for fragment coupling and introducing the 1,3-diketone moiety masked in the form of an isoxazole. We discovered that cycloaddition of the nitrile oxide and the enyne gave the rare 3,4-disubstituted isoxazole adduct under kinetic reaction conditions. The cycloaddition was found to be reversible, and the thermodynamic 3,5-disubstituted isoxazole could be obtained by isomerization of its 3,4-disubstituted isomer under elevated temperature. Our mechanistic studies support the role of hydrogen bonding in accelerating the isomerization. Consistent with our previous studies, the Z,E-macrocyclic bis-enone was found to be inactive toward the transannular bis-Michael reaction under the conditions evaluated.

Osteo-pathological analysis provides evidence for a survived historical ship strike in a Southern Hemisphere fin whale (Balaenoptera physalus).

The life history of a fin whale (Balaenoptera physalus) caught during whaling operations in the 1950s was partly reconstructed. 3D surface models of the bones of the skeleton curated at the Zoological Museum of Hamburg were used for an osteopathological analysis. The skeleton revealed multiple healed fractures of ribs and a scapula. Moreover, the processus spinosi of several vertebrae were deformed and arthrosis was found. Together, the pathological findings provide evidence for large blunt trauma and secondary effects arising from it. Reconstruction of the likely cause of events suggests collision with a ship inflicting the fractures and leading to post traumatic posture damage as indicated by skeletal deformations. The injured bones had fully healed before the fin whale was killed by a whaler in the South Atlantic in 1952. This study is the first in-detail reconstruction of a historical whale-ship collision in the Southern Hemisphere, dating back to the 1940s, and the first documentation of a healed scapula fracture in a fin whale. The skeleton provides evidence for survival of a ship strike by a fin whale with severe injuries causing long-term impairment.

Evidence of correlated evolution of hypsodonty and exceptional longevity in endemic insular mammals.

Here, we test whether the increase in tooth height in insular endemics results from the expansion of the dietary niche under resource limitation, as widely considered, or whether it represents an investment in dental durability in response to the selection for extended longevity under low levels of extrinsic mortality. We tested these hypotheses in the extremely hypsodont fossil bovid Myotragus balearicus from the Balearic Islands, an ideal model to study the evolutionary trends on islands. Dental abrasion was significantly lower in the insular bovid than in highly hypsodont continental artiodactyls, suggesting that feeding habits are not the sole driving force behind increased crown height. However, the estimated longevity for M. balearicus based on dental durability was two times that predicted from body mass. Survivorship curves confirm that an extraordinarily large number of individuals approached the longevity of the species. Our results, hence, provide evidence that hypsodonty in insular endemics is the outcome of selection for increased durability of the permanent dentition in association with an extended lifespan. In the context of insularity, our results lend additional support to the disposable soma theory of ageing confirming the dependency of somatic maintenance and repair on lifespan, and its control by resource availability and extrinsic mortality.

Teasing apart the contributions of hard dietary items on 3D dental microtextures in primates.

3D dental microtexture analysis is a powerful tool for reconstructing the diets of extinct primates. This method is based on the comparison of fossils with extant species of known diet. The diets of primates are highly diversified and include fruits, seeds, grass, tree leaves, bark, roots, tubers, and animal resources. Fruits remain the main component in the diets of most primates. We tested whether the proportion of fruit consumed is correlated with dental microtexture. Two methods of microtexture analysis, the scale-sensitive fractal analysis (SSFA) and the Dental Areal Surface Texture Analysis (DASTA; after ISO/FDIS 25178-2), were applied to specimens of eight primate species (Alouatta seniculus, Gorilla gorilla, Lophocebus albigena, Macaca fascicularis, Pan troglodytes, Papio cynocephalus, Pongo abelii, Theropithecus gelada). These species largely differ in the mean annual proportion of fruit (from 0 to 90%) in their diet, as well as in their consumption of other hard items (seeds, bark, and insect cuticles) and of abrasive plants. We find the complexity and heterogeneity of textures (SSFA) to correlate with the proportion of fruits consumed. Textural fill volume (SSFA) indicates the proportion of both fruits and other hard items processed. Furthermore, anisotropy (SSFA) relates to the consumption of abrasive plants like grass and other monocots. ISO parameters valley height, root mean square height, material volume, density of peaks, and closed hill and dale areas (DASTA) describe the functional interaction between food items and enamel facets during mastication. The shallow, plastic deformation of enamel surfaces induced by small hard particles, such as phytoliths or dust, results in flat microtexture relief, whereas the brittle, deep fracture caused by large hard items such as hard seeds creates larger relief.

Case report: surviving a tiger attack.

Attacks on humans by large predators are rare, especially in Northern Europe. In cases of involvement of the craniocervical compartment, most of the attacks are not survived. We report on a case where the patient survived a tiger attack despite severe head trauma and discuss the circumstances leading to the patient's survival and excellent outcome. The patient we report on is a 28-year-old tamer, who was attacked by three tigers during an evening show. A bite to the head resulted in multiple injuries including left-sided skull penetration wounds with dislocated fractures, dural perforations, and brain parenchyma lesions. The patient recovered without neurological deficits after initial ICU treatment. No infection occurred. In order to understand the mechanism of the tiger's bite to the patient's cranium, a simulation of the attack was performed using a human and a tiger skull put together at identical positions to the bite marks in a CT scan. It seems that during the bite, the animal was not able to clamp down on the patient's skull between its canine teeth and therefore reduced bite forces were applied. Survival of an attack by a large predator that targeted the cervical-cranial compartment with an excellent outcome is not described in the literature. We were surprised to find only minor lesions of the brain parenchyma despite the obvious penetration of the skull by the tiger's canines. This seems to be related to the specific dynamics of the cranial assault and the reduced forces applied to the patient's head demonstrated in a 3D bite simulation.

Post-mortem enamel surface texture alteration during taphonomic processes-do experimental approaches reflect natural phenomena?

Experimental approaches are often used to better understand the mechanisms behind and consequences of post-mortem alteration on proxies for diet reconstruction. Dental microwear texture analysis (DMTA) is such a dietary proxy, using dental wear features in extant and extinct taxa to reconstruct feeding behaviour and mechanical food properties. In fossil specimens especially, DMTA can be biased by post-mortem alteration caused by mechanical or chemical alteration of the enamel surface. Here we performed three different dental surface alteration experiments to assess the effect of common taphonomic processes by simplifying them: (1) tumbling in sediment suspension to simulate fluvial transport, (2) sandblasting to simulate mechanical erosion due to aeolian sediment transport, (3) acid etching to simulate chemical dissolution by stomach acid. For tumbling (1) we found alteration to be mainly dependent on sediment grain size fraction and that on specimens tumbled with sand fractions mainly post-mortem scratches formed on the dental surface, while specimens tumbled with a fine-gravel fraction showed post-mortem formed dales. Sandblasting (2) with loess caused only negligible alteration, however blasting with fine sand quartz particles resulted in significant destruction of enamel surfaces and formation of large post-mortem dales. Acid etching (3) using diluted hydrochloric acid solutions in concentrations similar to that of predator stomachs led to a complete etching of the whole dental surface, which did not resemble those of teeth recovered from owl pellets. The experiments resulted in post-mortem alteration comparable, but not identical to naturally occurring post-mortem alteration features. Nevertheless, this study serves as a first assessment and step towards further, more refined taphonomic experiments evaluating post-mortem alteration of dental microwear texture (DMT).

Another one bites the dust: faecal silica levels in large herbivores correlate with high-crowned teeth.

The circumstances of the evolution of hypsodonty (= high-crowned teeth) are a bone of contention. Hypsodonty is usually linked to diet abrasiveness, either from siliceous phytoliths (monocotyledons) or from grit (dusty environments). However, any empirical quantitative approach testing the relation of ingested silica and hypsodonty is lacking. In this study, faecal silica content was quantified as acid detergent insoluble ash and used as proxy for silica ingested by large African herbivores of different digestive types, feeding strategies and hypsodonty levels. Separate sample sets were used for the dry (n = 15 species) and wet (n = 13 species) season. Average faecal silica contents were 17-46 g kg(-1) dry matter (DM) for browsing and 52-163 g kg(-1) DM for grazing herbivores. No difference was detected between the wet (97.5 ± 14.4 g kg(-1) DM) and dry season (93.5 ± 13.7 g kg(-1) DM) faecal silica. In a phylogenetically controlled analysis, a strong positive correlation (dry season r = 0.80, p < 0.0005; wet season r = 0.74, p < 0.005) was found between hypsodonty index and faecal silica levels. While surprisingly our results do not indicate major seasonal changes in silica ingested, the correlation of faecal silica and hypsodonty supports a scenario of a dominant role of abrasive silica in the evolution of high-crowned teeth.