Scientific discovery in the age of artificial intelligence.

Artificial intelligence (AI) is being increasingly integrated into scientific discovery to augment and accelerate research, helping scientists to generate hypotheses, design experiments, collect and interpret large datasets, and gain insights that might not have been possible using traditional scientific methods alone. Here we examine breakthroughs over the past decade that include self-supervised learning, which allows models to be trained on vast amounts of unlabelled data, and geometric deep learning, which leverages knowledge about the structure of scientific data to enhance model accuracy and efficiency. Generative AI methods can create designs, such as small-molecule drugs and proteins, by analysing diverse data modalities, including images and sequences. We discuss how these methods can help scientists throughout the scientific process and the central issues that remain despite such advances. Both developers and users of AI toolsneed a better understanding of when such approaches need improvement, and challenges posed by poor data quality and stewardship remain. These issues cut across scientific disciplines and require developing foundational algorithmic approaches that can contribute to scientific understanding or acquire it autonomously, making them critical areas of focus for AI innovation.

Large-scale foundation model on single-cell transcriptomics.

Large pretrained models have become foundation models leading to breakthroughs in natural language processing and related fields. Developing foundation models for deciphering the 'languages' of cells and facilitating biomedical research is promising yet challenging. Here we developed a large pretrained model scFoundation, also named 'xTrimoscFoundationα', with 100 million parameters covering about 20,000 genes, pretrained on over 50 million human single-cell transcriptomic profiles. scFoundation is a large-scale model in terms of the size of trainable parameters, dimensionality of genes and volume of training data. Its asymmetric transformer-like architecture and pretraining task design empower effectively capturing complex context relations among genes in a variety of cell types and states. Experiments showed its merit as a foundation model that achieved state-of-the-art performances in a diverse array of single-cell analysis tasks such as gene expression enhancement, tissue drug response prediction, single-cell drug response classification, single-cell perturbation prediction, cell type annotation and gene module inference.

Progress and opportunities of foundation models in bioinformatics.

Bioinformatics has undergone a paradigm shift in artificial intelligence (AI), particularly through foundation models (FMs), which address longstanding challenges in bioinformatics such as limited annotated data and data noise. These AI techniques have demonstrated remarkable efficacy across various downstream validation tasks, effectively representing diverse biological entities and heralding a new era in computational biology. The primary goal of this survey is to conduct a general investigation and summary of FMs in bioinformatics, tracing their evolutionary trajectory, current research landscape, and methodological frameworks. Our primary focus is on elucidating the application of FMs to specific biological problems, offering insights to guide the research community in choosing appropriate FMs for tasks like sequence analysis, structure prediction, and function annotation. Each section delves into the intricacies of the targeted challenges, contrasting the architectures and advancements of FMs with conventional methods and showcasing their utility across different biological domains. Further, this review scrutinizes the hurdles and constraints encountered by FMs in biology, including issues of data noise, model interpretability, and potential biases. This analysis provides a theoretical groundwork for understanding the circumstances under which certain FMs may exhibit suboptimal performance. Lastly, we outline prospective pathways and methodologies for the future development of FMs in biological research, facilitating ongoing innovation in the field. This comprehensive examination not only serves as an academic reference but also as a roadmap for forthcoming explorations and applications of FMs in biology.

Assessing protein model quality based on deep graph coupled networks using protein language model.

Model quality evaluation is a crucial part of protein structural biology. How to distinguish high-quality models from low-quality models, and to assess which high-quality models have relatively incorrect regions for improvement, are remain a challenge. More importantly, the quality assessment of multimer models is a hot topic for structure prediction. In this study, we propose GraphCPLMQA, a novel approach for evaluating residue-level model quality that combines graph coupled networks and embeddings from protein language models. The GraphCPLMQA consists of a graph encoding module and a transform-based convolutional decoding module. In encoding module, the underlying relational representations of sequence and high-dimensional geometry structure are extracted by protein language models with Evolutionary Scale Modeling. In decoding module, the mapping connection between structure and quality is inferred by the representations and low-dimensional features. Specifically, the triangular location and residue level contact order features are designed to enhance the association between the local structure and the overall topology. Experimental results demonstrate that GraphCPLMQA using single-sequence embedding achieves the best performance compared with the CASP15 residue-level interface evaluation methods among 9108 models in the local residue interface test set of CASP15 multimers. In CAMEO blind test (20 May 2022 to 13 August 2022), GraphCPLMQA ranked first compared with other servers (https://www.cameo3d.org/quality-estimation). GraphCPLMQA also outperforms state-of-the-art methods on 19, 035 models in CASP13 and CASP14 monomer test set.

AttentionPert: accurately modeling multiplexed genetic perturbations with multi-scale effects.

MOTIVATION: Genetic perturbations (e.g. knockouts, variants) have laid the foundation for our understanding of many diseases, implicating pathogenic mechanisms and indicating therapeutic targets. However, experimental assays are fundamentally limited by the number of measurable perturbations. Computational methods can fill this gap by predicting perturbation effects under novel conditions, but accurately predicting the transcriptional responses of cells to unseen perturbations remains a significant challenge. RESULTS: We address this by developing a novel attention-based neural network, AttentionPert, which accurately predicts gene expression under multiplexed perturbations and generalizes to unseen conditions. AttentionPert integrates global and local effects in a multi-scale model, representing both the nonuniform system-wide impact of the genetic perturbation and the localized disturbance in a network of gene-gene similarities, enhancing its ability to predict nuanced transcriptional responses to both single and multi-gene perturbations. In comprehensive experiments, AttentionPert demonstrates superior performance across multiple datasets outperforming the state-of-the-art method in predicting differential gene expressions and revealing novel gene regulations. AttentionPert marks a significant improvement over current methods, particularly in handling the diversity of gene perturbations and in predicting out-of-distribution scenarios. AVAILABILITY AND IMPLEMENTATION: Code is available at https://github.com/BaiDing1234/AttentionPert.

Exploration of Sweeping Effect: Droplet Coalescence Jumping of a Rolling and Static Droplet.

The sweeping effect of merged droplets plays a key role in enhancing application performance due to the continuing coalescence caused by the horizontal jumping velocity. Most studies focused on static droplet coalescence jumping, while moving droplet coalescence is poorly understood. In this work, we experimentally and numerically study the coalescence of a rolling droplet and a static one. When the droplet radius ratio is larger than 0.8, as the dimensionless initial velocity increases and the vertical jumping velocity first decreases and then increases. The critical dimensionless initial velocity Vc* corresponding to the minimum vertical jumping velocity could be estimated as 0.9(rs2rm2). When the droplet radius ratio is smaller than 0.8, the dimensionless initial velocity has a positive effect on the vertical jumping velocity. The mechanism of the vertical jumping velocity can be attributed to two parts: liquid bridge impact and retraction of the merged droplet. The squeezing effect generated by the initial velocity between the two droplets promotes the growth of the liquid bridge and enhances the impact effect of the liquid bridge but weakens the upward velocity accumulation caused by the retraction of the merged droplets. However, different from the vertical jumping velocity, the horizontal jumping velocity is approximately proportional to the dimensionless initial velocity. The outcome of our work elucidates a fundamental understanding of a rolling droplet coalescing with a static one.

An isothermal CRISPR- based lateral flow assay for detection of Neisseria meningitidis.

BACKGROUND: Neisseria meningitidis can cause life-threatening meningococcal meningitis and meningococcemia. Old standard microbiological results from CSF/blood cultures are time consuming. This study aimed to combine the sensitivity of loop-mediated isothermal nucleic acid amplification (LAMP) with the specificity of CRISPR/Cas12a cleavage to demonstrate a reliable diagnostic assay for rapid detection of N. meningitidis. METHODS: A total of n = 139 samples were collected from patients with suspected meningococcal disease and were used for evaluation. The extracted DNA was subjected to qualitative real-time PCR, targeting capsular transporter gene (ctrA) of N. meningitidis. LAMP-specific primer pairs, also targeting the ctrA, were designed and the LAMP products were subjected to CRISPR/Cas12 cleavage reaction. the readout was on a lateral flow strip. Sensitivity, specificity, positive predictive value (PPV) and negative predictive value (NPV) of LAMP-CRISPR/Cas was compared with real-time PCR assays. The limit of detection (LOD) was established with serial dilutions of the target N. meningitidis DNA and calculated by Probit regression analysis. RESULTS: Six LAMP assay-specific primers were developed targeting the ctrA gene of N. meningitidis, which is conserved in all meningococcal serogroups. The LAMP primers did not amplify DNA from other bacterial DNA tested, showing 100% specificity. The use of 0.4 M betaine increased the sensitivity and stability of the reaction. LAMP-CRISPR/Cas detected meningococcal serogroups (B, C, W). The assay showed no cross-reactivity and was specific for N. meningitidis. The LOD was 74 (95% CI: 47-311) N. meningitidis copies. The LAMP-CRISPR/Cas performed well compared to the gold standard. In the 139 samples from suspected patients, the sensitivity and specificity of the test were 91% and 99% respectively. CONCLUSION: This developed and optimized method can complement for the available gold standard for the timely diagnosis of meningococcal meningitis and meningococcemia.

Supercritical CO2 fluid extract from Stellariae Radix ameliorates 2,4-dinitrochlorobenzene-induced atopic dermatitis by inhibit M1 macrophages polarization via AMPK activation.

Yin chai hu (Radix Stellariae) is a root medicine that is frequently used in Chinese traditional medicine to treat fever and malnutrition. In modern medicine, it has been discovered to have anti-inflammatory, anti-allergic, and anticancer properties. In a previous study, we were able to extract lipids from Stellariae Radix using supercritical CO2 extraction (SRE), and these sterol lipids accounted for up to 88.29% of the extract. However, the impact of SRE on the development of atopic dermatitis (AD) has not yet been investigated. This study investigates the inhibitory effects of SRE on AD development using a 2,4-dinitrochlorobenzene (DNCB)-induced AD mouse model. Treatment with SRE significantly reduced the dermatitis score and histopathological changes compared with the DNCB group. The study found that treatment with SRE resulted in a decrease of pro-inflammatory cytokines TNF-α, CXC-10, IL-12, and IL-1ß in skin lesions. Additionally, immunohistochemical analysis revealed that SRE effectively suppressed M1 macrophage infiltration into the AD lesion. Furthermore, the anti-inflammatory effect of SRE was evaluated in LPS + INF-γ induced bone marrow-derived macrophages (BMDMs) M1 polarization, SRE inhibited the production of TNF-α, CXC-10, IL-12, and IL-1ß and decreased the expression of NLRP3. Additionally, SRE was found to increase p-AMPKT172, but had no effect on total AMPK expression, after administration of the AMPK inhibitor Compound C, the inhibitory effect of SRE on M1 macrophages was partially reversed. The results indicate that SRE has an inhibitory effect on AD, making it a potential therapeutic agent for this atopic disorder.

Prevalence and Genetic Diversity of Entamoeba and Cryptosporidium in Pigs and Wild Boars in Central and Southern Vietnam: Implications for Zoonotic Risks and Surveillance.

Background: Parasites of Entamoeba and Cryptosporidium genera, prevalent among various vertebrates such as humans and pigs, pose a zoonotic threat as common protozoan pathogens. This study investigated the prevalence and genetic diversity of Entamoeba and Cryptosporidium species in pigs and wild boars across central and southern Vietnam, to ascertain parasite transmission dynamics. Methods: A total of 113 independent stool samples from 77 pigs and 36 wild boars were analyzed using PCR-based molecular methodologies to detect the presence of Entamoeba spp. and Cryptosporidium spp. The identified species were further characterized through Sanger sequencing, and phylogenetic relationships were analyzed. Results: The study revealed a high prevalence of Entamoeba spp. (62%, n = 70/113) and Cryptosporidium spp. (31%, n = 35/113). Entamoeba suis (57%, n = 40) was predominant, followed by Entamoeba polecki (40%, n = 40) and Entamoeba hartmanni (3%, n = 2). Among Cryptosporidium species, Cryptosporidium scrofarum (89%, n = 31) was the most common, followed by Cryptosporidium suis (11%, n = 4). Wild boars exhibited a higher prevalence of Entamoeba infection compared with domestic pigs (p = 0.019). Conclusions: The study highlights a high prevalence of Entamoeba and Cryptosporidium, suggesting a potential for zoonotic transmission in Vietnam. Further investigations are necessary to determine the extent to which these parasites in pigs and wild boars contribute to the burden in the human population.

Physiological, Metabolome and Gene Expression Analyses Reveal the Accumulation and Biosynthesis Pathways of Soluble Sugars and Amino Acids in Sweet Sorghum under Osmotic Stresses.

Water scarcity is a major environmental constraint on plant growth in arid regions. Soluble sugars and amino acids are essential osmolytes for plants to cope with osmotic stresses. Sweet sorghum is an important bioenergy crop and forage with strong adaptabilities to adverse environments; however, the accumulation pattern and biosynthesis basis of soluble sugars and amino acids in this species under osmotic stresses remain elusive. Here, we investigated the physiological responses of a sweet sorghum cultivar to PEG-induced osmotic stresses, analyzed differentially accumulated soluble sugars and amino acids after 20% PEG treatment using metabolome profiling, and identified key genes involved in the biosynthesis pathways of soluble sugars and amino acids using transcriptome sequencing. The results showed that the growth and photosynthesis of sweet sorghum seedlings were significantly inhibited by more than 20% PEG. After PEG treatments, the leaf osmotic adjustment ability was strengthened, while the contents of major inorganic osmolytes, including K+ and NO3-, remained stable. After 20% PEG treatment, a total of 119 and 188 differentially accumulated metabolites were identified in the stems and leaves, respectively, and the accumulations of soluble sugars such as raffinose, trehalose, glucose, sucrose, and melibiose, as well as amino acids such as proline, leucine, valine, serine, and arginine were significantly increased, suggesting that these metabolites should play key roles in osmotic adjustment of sweet sorghum. The transcriptome sequencing identified 1711 and 4978 DEGs in the stems, as well as 2061 and 6596 DEGs in the leaves after 20% PEG treatment for 6 and 48 h, respectively, among which the expressions of genes involved in biosynthesis pathways of sucrose (such as SUS1, SUS2, etc.), trehalose (including TPS6), raffinose (such as RAFS2 and GOLS2, etc.), proline (such as P5CS2 and P5CR), leucine and valine (including BCAT2), and arginine (such as ASS and ASL) were significantly upregulated. These genes should be responsible for the large accumulation of soluble sugars and amino acids under osmotic stresses. This study deepens our understanding of the important roles of individual soluble sugars and amino acids in the adaptation of sweet sorghum to water scarcity.

Vision-Based Contact Adhesion Measurement Method on Soft Matter Surfaces.

Adhesion property measurements contribute to a comprehensive understanding of the mechanical properties of soft matters. Indentation tests are a common method for measuring the adhesion force. However, indenters generally have a large volume and a small sensing angle and, thus, are not conducive to local detection in high-precision environments. Here, we propose a vision-based contact adhesion measurement (VisCAM) method to achieve the contact image and adhesion force on soft matter surfaces from the perspective of indentation direction. The coupling of the 7.6 mm diameter probe and a flexible fiber makes the system similar to a miniaturized endoscope. Classical contact theories and finite element models are used for the contact mechanics analysis of silicone rubber. The image grayscale-load mathematical model is constructed based on the change in contact light spot. Finally, the uncertainty of the system is less than 4%, and the measurement error is 0.04 N. In-vitro kidney indentation experiments showed that the local adhesion force measurement of soft tissues can be completed. Our method provides better solutions for understanding the adhesion properties of soft matters.

Exploring the Effects of Liquid Marbles' Deformation on Their Rolling Resistance.

Liquid marbles (LMs) are nonwetting droplets manufactured by encapsulating droplets with micro- or nanoscale particles. These marbles are widely used as transport carriers for digital microfluidics due to their rapid displacement velocity and leak-free transport. An improved understanding of the resistance mechanism of rolling LMs is crucial for their transport and manipulation. In this study, we investigated the rolling resistance of LMs obtained with different powders and volumes using a high-speed camera. Our findings suggest that the deformation of liquid marbles would hinder their rolling by a resistance torque. To depict this resistance effect, we propose a theoretical model (f∼λ(ε-12Bo1/2ε2+14Boε3)), where f is the rolling resistance of marbles, λ is the deflection coefficient, Bo is the Bond number, and (ε is the contact surface deformation) that accurately predicts the relationship between deformation and rolling resistance, which is supported by our experimental results. To further validate our theoretical model, we conducted three independent experiments: shape detection of prepared LMs, measuring the elastic force of LMs, and detecting the diffusive motion of the encapsulating particles. Furthermore, we discuss three factors that affect the rolling resistance: the volume of the marbles, the type and size of the encapsulating particles, and the substrate roughness. This comprehensive study not only generalizes the mechanism of deformation hindering the rolling of liquid marbles but also provides a theoretical framework to predict the relationship between the deformation and rolling resistance. These findings have practical implications for improving the manipulation efficiency and advancing the use of LMs as microfluidic carriers.

An experimental study of the merging flow of polymer solutions in a T-shaped microchannel.

The merging flow through a T-junction is relevant to sample mixing and particle manipulation in microfluidic devices. It has been extensively studied for Newtonian fluids, particularly in the high inertial regime where flow bifurcation takes place for enhanced mixing. However, the effects of fluid rheological properties on the merging flow have remained largely unexplored. We investigate here the flow of five types of polymer solutions along with water in a planar T-shaped microchannel over a wide range of flow rates for a systematic understanding of the effects of fluid shear thinning and elasticity. It is found that the merging flow near the stagnation point of the T-junction can either be vortex dominated or have unsteady streamlines, depending on the strength of elasticity and shear thinning present in the fluid. Moreover, the shear thinning effect is found to induce a symmetric unsteady flow in comparison to the asymmetric unsteady flow in the viscoelastic fluids, the latter of which exhibits greater interfacial fluctuations.

Association Between Soluble Notch Ligand Delta-like Ligand 1 and Bleeding Complications in Patients With Dengue Fever Infection.

Bleeding associated with endothelial damage is a key feature of severe dengue fever. In the current study, we investigated whether Notch ligands were associated with bleeding in 115 patients with confirmed dengue infection in Vietnam. Soluble Notch ligands were determined by means of enzyme-linked immunosorbent assay. Seventeen of 115 patients (14.8%) experienced bleeding manifestations. High soluble delta-like ligand 1 (sDLL1) plasma levels was associated with bleeding (median, 15 674 vs 7117 pg/mL; P < .001). Receiver operating characteristic (ROC) curve analysis demonstrated that sDLL1 had the best test performance (area under the ROC curve, 0.852), with 88% sensitivity and 84% specificity. The combination with alanine aminotransferase and aspartate aminotransferase slightly increased sDLL1 performance. sDLL1 may be useful to guide clinical management of patients with patients in endemic settings.

Laplace Pressure Difference Enhances Droplet Coalescence Jumping on Superhydrophobic Structures.

Coalescence-induced droplet jumping has great prospects in many applications. Nevertheless, the applications are vastly limited by a low jumping velocity. Conventional methods to enhance the droplet coalescence jumping velocity are enabled by protruding structures with superhydrophobic surfaces. However, the jumping velocity improvement is limited by the height of protruding structures. Here, we present rationally designed limitation structures with superhydrophobic surfaces to achieve a dimensionless jumping velocity, Vj* ≈ 0.64. The mechanism of enhancing the jumping velocity is demonstrated through the study of numerical simulations and geometric parameters of limitation structures, providing guidelines for optimized structures. Experimental and numerical results indicate that the mechanism consists of the combined action of the velocity vectors' redirection and the Laplace pressure difference within deformed droplets trapped in limitation structures. On the basis of previous research on the mechanisms of protruding structures and our study, we successfully exploited those mechanisms to further improve the jumping velocity by combining the limitation structure with the protruding structure. Experimentally, we attained a dimensionless jumping velocity of Vj* ≈ 0.74 with an energy conversion efficiency of η ≈ 48%, breaking the jumping velocity limit. This work not only demonstrates a new mechanism for achieving a high jumping velocity and energy conversion efficiency but also sheds lights on the effect of limitation structures on coalescence hydrodynamics and elucidates a method to further enhance the jumping velocity based on protruding structures.

CRISPR-Cas12a combination to alleviate the false-positive in loop-mediated isothermal amplification-based diagnosis of Neisseria meningitidis.

BACKGROUND: Loop isothermal amplification (LAMP) has recently been proposed as a point-of-care diagnostic tool to detect acute infectious pathogens; however, this technique embeds risk of generating false-positive results. Whereas, with abilities to accurately recognize specific sequence, the CRISPR/Cas12a can forms complexes with cognate RNA sensors and cleave pathogen's DNA targets complimerntary to its cognate RNA, afterward acquiring the collateral activity to unbiasedly cut nearby off-target fragments. Therefore, if relevant fluorescent-quencher-nucleic probes are present in the reaction, the non-specific cleavage of probes releases fluorescences and establish diagnostic read-outs. METHODS: The MetA gene of N. meningitidis was selected as target to optimize the LAMP reaction, whereas pseudo-dilution series of N. meningitidis gemonics DNA was used to establish the detection limit of LAMP/Cas12a combination assay. The diagnostic performance of established LAMP/Cas12a combination assay was validated in comparation with standard real-time PCR on 51 CSF samples (14 N. meningitidis confirmed patients and 37 control subjects). RESULTS: In relevant biochemical conditions, CRISPR-Cas12a and LAMP can work synchronously to accurately identify genetics materials of Nesseria menitigistis at the level 40 copies/reaction less than 2 h. CONCLUSIONS: In properly optimized conditions, the CRISPR-Cas12a system helps to alleviate false positive result hence enhancing the specificity of the LAMP assays.

Diagnosis of pathogens causing bacterial meningitis using Nanopore sequencing in a resource-limited setting.

AIM: The aim of the present study is to compare the performance of 16S rRNA Nanopore sequencing and conventional culture in detecting infectious pathogens in patients with suspected meningitis in a resource-limited setting without extensive bioinformatics expertise. METHODS: DNA was isolated from the cerebrospinal fluid (CSF) of 30 patients with suspected bacterial meningitis. The isolated DNA was subjected to 16S sequencing using MinION™. The data were analysed in real time via the EPI2ME cloud platform. The Nanopore sequencing was done in parallel to routine microbiological diagnostics. RESULTS: Nanopore sequencing detected bacterial pathogens to species level in 13 of 30 (43%) samples. CSF culture showed 40% (12/30) positivity. In 21 of 30 patients (70%) with suspected bacterial meningitis, both methods yielded concordant results. About nine of 30 samples showed discordant results, of these five were false positive and four were false negative. In five of the culture negative results, nanopore sequencing was able to detect pathogen genome, due to the higher sensitivity of the molecular diagnostics. In two other samples, the CSF culture revealed Cryptococcus neoformans and Streptococcus pneumoniae, which were not detected by Nanopore sequencing. Overall, using both the cultures and 16S Nanopore sequencing, positivity rate increased from 40% (12/30) to 57% (17/30). CONCLUSION: Next-generation sequencing could detect pathogens within six hours and could become an important tool for both pathogen screening and surveillance in low- and middle-income countries (LMICs) that do not have direct access to extensive bioinformatics expertise.

Pharmacological Therapies for Osteoporosis: A Bayesian Network Meta-Analysis.

BACKGROUND Numerous randomized controlled trials (RCTs) have evaluated pharmacological therapies for osteoporosis. The aim of this Bayesian network meta-analysis was to compare the efficacy and safety of pharmacological therapies for osteoporosis patients. MATERIAL AND METHODS The electronic databases of PubMed, Embase, and Cochrane Library were systematically searched for eligible RCTs from their inception up to January 2021. The primary endpoints were all fractures, vertebral fractures, and non-vertebral fractures, while the secondary endpoints were fractures at hip or peripheral locations, bone mineral density (BMD) at various sites, and potential adverse events. RESULTS We included 79 RCTs reporting a total of 108 797 individuals in the final quantitative analysis. The results of network analysis indicated that romosozumab (92.1%) was the most effective in reducing the risk for all fractures, with the best therapeutic effects on vertebral fracture (97.2%) and non-vertebral fracture (88.0%). Romosozumab (92.5%) provided better therapeutic effects for the reduction of hip fracture. The best treatment agents for improving whole-body BMD (100.0%), spine BMD (95.7%), hip BMD (92.4%), femoral neck BMD (86.7%), and trochanter BMD (95.5%) were alendronate, strontium ranelate, ibandronate, risedronate, and ibandronate, respectively. Finally, the use of bazedoxifene was associated with the highest incidence of any upper-gastrointestinal event, nasopharyngitis, and back pain, while risedronate was associated with higher incidence of abdominal pain and dyspepsia. CONCLUSIONS This study found that romosozumab yielded the best effects for preventing fracture risk, while abaloparatide was the most effective in reducing the risk of vertebral fracture and non-vertebral fracture.

High-precision and long-range measurement of micro-gear starting torque.

The ability to measure micro-starting torque is pivotal for micromechanical equipment, which has wide usage in mechanical manufacturing, electrical, electronic, and other industries. However, the measurement range of existing methods is about N⋅m or mN⋅m. There is not much research on the measurement of micro-torque starting in the µN⋅m. In this paper, a novel micro-gear starting torque measurement system, to the best of our knowledge, is proposed based on an optical lever with a long range from 1 to 10µN⋅m. The system device consists of the optical lever, cantilever, and position sensitive device. A micro-gear was used to assess the performance of the proposed method. The standard deviation of the measured starting torque is 1.2µN⋅m. The external factors that can contribute to the uncertainty of the measurement system, such as force measurement, arm of force, and repeatability, have been analyzed and quantified. The relative combined uncertainty is estimated at 3.0%, approximately.