Biology

• Integrated three public experimental tables and cleaned 11,696 promoter–RBS combinations. Reconstructed RNA sequences around the protein translation-initiation region using measured transcription start sites, computed RNA structural accessibility, and used XGBoost with 12 sequence and structural features to predict relative protein expression (a fluorescence measure).
• To evaluate unseen elements, held out promoters, RBSs, and both together, separating element identities between training and testing. Used five-fold validation, compared with an XGBoost model using nine sequence-only features under the same conditions, and estimated differences with 1,000 grouped bootstrap iterations.
• Adding structural features reduced log mean absolute error (log-MAE) for the protein/RNA ratio, a translation-related proxy, by approximately 10%–13%. Improvement direction was consistent across four split seeds. These results come from data evaluation in a fixed reporter-protein context.
• Exposed nine tools for sequence analysis, expression prediction, and ranking through standard MCP interfaces for compatible AI platforms and clients. Given a promoter, ranked the 111 RBSs in the original data library by predicted expression.
• Designed 67 tests covering biological inputs, numerical calculations, evaluation isolation, and tool interfaces. Checked sequence coordinates, RNA structural energy, training/test element isolation, and prediction and candidate-ranking call chains to provide reproducible regression validation.

Promoter/RBS, TSS, RNA accessibility, ViennaRNA, XGBoost, OSTIR, grouped cross-validation, bootstrap, Python, MCP

• Cloning and validation: Participated in foundational NSE2 cloning and the group NSE3 extension, including target-fragment amplification, pRS414 restriction digestion and ligation, E. coli transformation, and initial blue-white screening. Checked constructs using restriction-fragment sizes; the initial NSE3 amplicon was approximately 1,920 bp.
• Marker insertion and mutagenesis: Inserted URA3 through overlap-extension PCR and used DpnI to remove the parental template. For NSE2 mutagenesis, compared amplification yield and mutagenesis intensity using a manganese-free control and a 160–480 μM MnCl2 gradient.
• Yeast screening: Transformed yeast using PEG/LiAc/single-stranded carrier DNA, with mutagenic, non-mutagenic, and control groups. Observed transformant growth on uracil-deficient medium and 1:10 dilution plates.
• Sequencing interpretation: Reviewed NSE3/URA3 constructs using Nanopore clone validation and amplicon reports. The barcode06 assembly measured 7,341 bp, with mean coverage of approximately 1,465× from 1,500 downsampled reads; organized sequence differences alongside construct-quality findings.

Molecular cloning, PCR, restriction digestion and ligation, overlap-extension PCR, mutagenic PCR, yeast transformation, Nanopore

• Immunization and observation: Emulsified supplied antigens with Freund’s adjuvant for immunization, tracked animal condition across multiple rounds, and promptly recorded and reported abnormalities.
• Samples and testing: Participated in blood collection and centrifugation to separate serum. Prepared samples for ELISA titer testing to support immunization assessment and subsequent delivery.

Antigen emulsification, laboratory-animal observation, blood-sample processing, serum separation, ELISA

• Ecological data analysis: Recorded vertebrate observations with an emphasis on birds. Compared species richness with historical data and plotted relative plant and animal abundance to describe community composition.
• Spatial presentation: Mapped observation locations onto a campus map using R leaflet, color-coding species distributions to connect statistical findings with their spatial context.

R, leaflet, species richness, relative abundance, spatial visualization

• Field survey: Participated in black stork population, distribution, and habitat surveys in suburban Beijing, recording potential threats and organizing observations.
• Data reporting: Reported survey data to the Beijing Municipal Forestry and Parks Bureau to support broader local bird-data compilation.

Field observation, species distribution, habitat survey, data reporting

• Data and QC: Selected GSE212598 and GSE203044 from GEO, totaling 12 RNA-seq samples, with three treated and three control samples per dataset. Used FastQC/MultiQC for quality checks and Kallisto for transcript quantification.
• Differential analysis: Applied the DESeq2 Wald test and BH/FDR to approximately 30,800 and 24,600 expression features. Using the existing adjusted p-value < 0.05 threshold, identified 11,675 and five features, respectively.
• Biological interpretation: Combined PCA, hierarchical clustering, and PANTHER/Reactome analyses of cell stress, cell cycle, and endoplasmic-reticulum stress responses. Connected expression separation in the luteolin data with pathway clues while distinguishing within-cohort conclusions from cross-cell-line comparisons.

RNA-seq, FastQC, MultiQC, Kallisto, DESeq2, R, BH/FDR, PCA, PANTHER, Reactome

• Pipeline practice: Documented a BWA-MEM → SAMtools → BCFtools → VEP workflow for public SRR077862 data, covering GRCh38 reference selection, attempted read alignment, alignment processing, and a variant-screening plan.
• Annotation and interpretation: Organized variant consequences from example VEP output, selected the OR4F5 K104Q missense change, and consulted dbSNP/UniProt and related literature. Combined amino-acid changes, population frequency, and functional predictions in the interpretation.
• Structural comparison: Combined the AlphaFold wild-type structure with a SWISS-MODEL variant-sequence model, located amino acid 104, and compared its local position and model confidence to produce structural visualizations and an analysis report.

BWA-MEM, SAMtools, BCFtools, VEP, GRCh38, dbSNP, UniProt, AlphaFold, SWISS-MODEL

• Target selection and in vitro checks: Used BLAST, CHOPCHOP, and Primer-BLAST to analyze candidate genes, sgRNAs, and primers. Performed PCR and in vitro Cas9 checks, examining experimental products using an approximately 503 bp optix amplicon and 2% agarose gels.
• Injection and phenotype records: Participated in embryo microinjection and developmental tracking. Archived adults by group, injection date, and individual identifier, arranged four wings consistently, photographed dorsal and ventral sides, and recorded color, symmetry, and eyespot features.
• Quantification and statistics: Used ImageJ to measure wing ROIs and black-area fractions, and R to analyze 241 observations aggregated by the class. Compared treatment and wild-type groups, using ventral eyespot observations to discuss the interpretive scope of a single area metric.

CRISPR/Cas9, BLAST, CHOPCHOP, Primer-BLAST, PCR, electrophoresis, microinjection, ImageJ, R