Discovery speed is advancing faster than proof of clinical success
THEME RESEARCH
Biotech
Will better biological models improve patient outcomes and portfolio economics, not just generate more candidates?
Multi-asset clinical attrition, patient efficacy, cycle time, and licensing economics
01TOP SIGNAL
Abstract screened
Read evidence →Personalized cancer vaccines are working, but manufacturing economics are the next test
The takeaway: Moderna and Merck have been working together for some years now on cancer therapies that would combine Merck’s Keytruda antibody with personalized mRNA immunotherapy from Moderna. And that “personalized” is no figure of speech: since each cancer patient’s tumor case is indeed somewhat different from any other, the idea is to use biopsy and blood samples to tailor the therapeutic-vaccine effects of the mRNA injections to each patient’s profile of neoantigens being produced by their tumors. That’s a real technological challenge - well, if you don’t plan on it costing ten million dollars for every patient, anyway. But your best shot is indeed with mRNA, because modern molecular biology techniques really allow you to manipulate it quickly and precisely. Moderna also has obvious experience on s
02
Breast cancer has a new protein target. Clinical proof is still missing
The model flags SUSD3 and the lab work shows it can promote cancer-cell growth and migration, but it has not shown that testing or targeting SUSD3 improves patient outcomes.
Full text reviewed
Europe PMC
03AI spotted a hidden liver cancer risk signal across multiple omics layers
The model combines several omics layers to classify liver-cancer risk, but small single-cell samples and reproducibility limits keep it out of clinical decision-making for now.
Full text reviewed
Europe PMC
04A smaller biology model beat bigger rivals by learning how genes connect
ScKGBERT mixes 41 million cell profiles with protein-interaction data, improving several biology benchmarks with fewer parameters; it still has not generated a prospectively validated drug target.
Full text reviewed
Europe PMC
05RNA may reveal missing protein data without another assay
ScProTrans predicts protein abundance from RNA and can estimate proteins not measured in the training assay, but predicted proteomics still needs experimental confirmation before researchers can rely on it.
Full text reviewed
Europe PMC
01TOP SIGNAL
Abstract screened
Read evidence →Enhanced antitumor immunity of mRNA vaccines by bioorthogonal-like delayed activation of exogeneous STING
The takeaway: Nature Biotechnology, Published online: 19 August 2026; doi:10.1038/s41587-026-03224-y mRNA vaccine efficacy is enhanced by delayed activation of a STING adjuvant.
02
AI antibody discovery finally faced a blind lab test
Nature Biotechnology, Published online: 19 August 2026; doi:10.1038/s41587-026-03238-6 A prospective, blinded competition is conducted for artificial-intelligence-generated and optimized antibody discovery.
Abstract screened
rss:www.nature.com
03High-throughput synthesis of DNA fragments by molecular self-assembly of overlapping oligonucleotides
Nature Biotechnology, Published online: 19 August 2026; doi:10.1038/s41587-026-03266-2 DNA libraries are synthesized at scale.
Source metadata only
rss:www.nature.com
04One pot can now build more than 1,000 gene fragments in parallel
Nature Biotechnology, Published online: 19 August 2026; doi:10.1038/s41587-026-03290-2 MOSAIC is a high-throughput gene synthesis method driven by hybridization (the self-assembly of complementary oligonucleotides during thermal annealing) instead of enzymatic processing. By using the full capacity of microchip-based oligonucleotide synthesis, MOSAIC enables the parallel production of over 1,000 distinct gene fragments and mutant libraries in a simple ‘one-pot’ reaction.
Abstract screened
rss:www.nature.com
05From reading to writing
Nature Biotechnology, Published online: 19 August 2026; doi:10.1038/s41587-026-03300-3 DNA writing is entering a period of rapid innovation, with enzymatic and cellular approaches challenging decades-old chemistry. Although none yet delivers long, accurate DNA at the scale biology ultimately requires, continued advances promise to remove one of the field’s most persistent bottlenecks and enable experiments that are currently out of reach.
Source metadata only
rss:www.nature.com
01TOP SIGNAL
Source metadata only
Read evidence →Cross-Species BAC Transgenesis Reveals Long-Range Regulation Drives Variation in Brain Oxytocin Receptor Expression and Social Behaviors
The takeaway: Although oxytocin (OXT) exhibits a highly conserved neuroanatomical pattern among vertebrates, the distribution of OXT receptor (OXTR) in brain varies considerably across species and is associated with species-typical social behavior. To investigate the genomic basis of the phylogenetic plasticity in brain Oxtr expression and its social behavioral consequences, we generated transgenic mice carrying a bacteria artificial chromosome (BAC) harboring the entire prairie vole Oxtr locus and flanking intergenic regulatory regions. We established eight independent "volized" mouse lines expressing prairie vole Oxtr (pvOxtr). Strikingly, despite conserved Oxtr expression in mammary gland of all transgenic mouse lines, each line displayed a unique pattern of brain expression distinct from both mice and prairie voles. Together with topologically associating domain (TAD) structure analysis with mouse genome, our findings suggest that unlike Oxt, Oxtr expression patterns in brain, involve contributions of distal regulatory elements beyond the BAC insert. In contrast, Oxtr expression in peripheral tissues appears resistant to such distal influences. Moreover, the "volized" mouse lines with different brain Oxtr expression patterns showed differences in partner preference and maternal behaviors, providing direct functional evidence that variation in brain Oxtr expression can drive differences in social behaviors. We propose that brain Oxtr expression is transcriptionally sensitive to long-range interactions with distal genomic elements, rendering it more susceptible to diverse regulatory influences. This supports a model in which regulatory flexibility facilitates the evolutionary diversification of social behavior, while maintaining essential peripheral Oxtr expression.
02
Cdc42 coordinates apical membrane formation and secretory granule maturation through v-SNARE maintenance in salivary acinar cells
Epithelial acinar cells of the salivary gland establish apical-basal polarity and secretory competence during development, processes that require tightly coordinated membrane trafficking. The Rho-family GTPase Cdc42 is a key regulator of epithelial polarity, but it has rarely been linked to the vesicle-fusion machinery in vivo. Using mice with acinar cell-specific deletion of Cdc42, we found that Cdc42 loss caused the accumulation of small vesicles near the apical surface, where the water channel AQP5 was retained in clusters and lumen formation was defective. Cdc42-deficient glands showed a selective reduction of two vesicle-associated SNAREs, VAMP2 and VAMP4, whereas VAMP8, the t-SNAREs, and an endocytic marker were unchanged. The corresponding mRNAs were not decreased, indicating that Cdc42 maintains these v-SNAREs through post-transcriptional mechanisms. Cdc42-deficient acinar cells also accumulated condensing vacuoles, a form of immature secretory granule, indicating impaired granule maturation. We propose that Cdc42 coordinates two distinct membrane-remodeling processes by maintaining these v-SNAREs: VAMP2 supports the fusion of AQP5-containing vesicles with the apical membrane, whereas VAMP4 supports secretory granule maturation. These findings identify the vesicle-fusion machinery as a target of Cdc42-dependent polarity control, with potential implications for secretory gland disorders.
Source metadata only
biorxiv
03Lactic acid bacterium Fructilactobacillus sanfranciscensis impairs fitness of yeast Maudiozyma humilis in synthetic wheat sourdough
Sourdough starters contain simple microbial communities typically consisting of a few bacterial species and one or two yeast species. The yeast Maudiozyma humilis and the lactic acid bacterium Fructilactobacillus sanfranciscensis often co-occur in sourdough starters and have been presumed to exist in a trophic relationship supported by glucose cross-feeding. However, previous research has highlighted a lack of evidence showing that yeast strains consume the glucose that F. sanfranciscensis produces. We have investigated the interaction between sourdough isolates of M. humilis and F. sanfranciscensis in a synthetic wheat sourdough medium, allowing us to control substrate composition and use flow cytometry to enumerate living and dead cells. M. humilis fitness was found to be lower in co-culture with F. sanfranciscensis than when grown alone. Analysis of spent medium composition highlighted the reliance of M. humilis on glucose rather than maltose for growth. Comparisons of predicted and measured co-culture metabolite content also revealed that F. sanfranciscensis consumed less maltose in co-culture than when grown alone. For the first time, we examined potential amino acid cross-feeding between M. humilis and F. sanfranciscensis, and found that within the pairing, F. sanfranciscensis was the main producer of amino acids. Our findings suggest that the M. humilis-F. sanfranciscensis interaction is likely to be neutral, or even competitive, with the strain identity of F. sanfranciscensis playing a defining role in the observed dominance of the bacteria and spent medium metabolite composition. ImportanceThe association of the yeast Maudiozyma humilis and the bacterium Fructilactobacillus sanfranciscensis in sourdough starters is well-documented, and together this pairing makes key functional and organoleptic contributions to the final bread product. Their relationship has historically been thought to be stabilised by cross-feeding of glucose to M. humilis. However, this
Source metadata only
biorxiv
04Advances in the Design and Functionality of a Compact Multi-Reflecting Time-of-Flight Mass Spectrometer
AO_SCPLOWBSTRACTC_SCPLOWO_ST_ABSRationaleC_ST_ABSHigh-resolution mass spectrometry is routinely used for the analysis of complex samples in pharmaceutical, environmental, and omics related studies. Such applications demand instrumentation to be capable of combining sub-ppm mass accuracy, high resolving power, rapid full m/z range acquisition, over a wide dynamic range. MethodsAchieving the above requirements places constraints on analyzer design and performance. Multi-reflecting time-of-flight (MRT) based analyzers have previously been reported as a means of significantly extending the effective flight path in compact TOF designs. Here, further instrument and functionality advances in a compact MRT mass spectrometer design are described. ResultsThe impact of these enhancements was assessed for targeted and non-targeted omics applications, examining the impact of acquisition speed on resolving power, dynamic range including limits of quantitation, and quantitative precision. ConclusionThe results obtained characterize the performance of the enhanced design features of a compact MRT mass spectrometer. Operation at elevated acquisition rates up to 200 Hz was observed without loss in resolving power, isotopic ratio accuracy, or quantitative precision.
Source metadata only
biorxiv
05A longitudinal fNIRS hyperscanning dataset of same generation and intergenerational relationship development and collaboration
Social disconnection (i.e., loneliness and social isolation) poses a severe risk to health and is particularly prevalent among older adults. Qualitative and behavioural evidence suggests that intergenerational social interactions can mitigate social disconnection by offering opportunities for adults to form new social connections, yet the neurophysiological and kinematic processes underpinning relationship formation remain poorly understood. This article presents and validates the InterGenSynchrony Dataset -- a longitudinal, multimodal hyperscanning dataset acquired during a 6-session program, where 30 same generation and 31 intergenerational dyads met as strangers and became acquainted through creative drawing. The dataset includes dyads' concurrent recordings of brain activity (i.e., fNIRS hyperscanning), motion capture, self-report measures including loneliness, collaborative behavioural scores, drawings (artefacts of collaboration), subjective experiences in text and interview forms. This dataset is optimised for multimodal investigations into the neural, physiological, behavioural, kinematic, and subjective aspects of relationship formation within and between generations in a real-world setting.
Source metadata only
biorxiv