- M.S., Computer Engineering (Bioinformatics) | Sharif University of Technology (December 2023)
- B.S., Computer Engineering | Semnan University (December 2020)
- The application of deep learning in Bioinformatics, NLP, and text analysis
- Machine Learning and Deep Learning approach
- Data science, especially Big Data analysis
1- An optimized graph-based structure for single-cell RNA-seq cell-type classification based on non-linear dimension reduction
It is now possible to analyze cellular heterogeneity at the single-cell level thanks to the rapid developments in single-cell sequencing technologies. The clustering of cells is a fundamental and common step in heterogeneity analysis. Even so, accurate cell clustering remains a challenge due to the high levels of noise, the high dimensions, and the high sparsity of data. we present SCEA, a clustering approach for scRNA-seq data. Using two consecutive units, an encoder based on MLP and a graph attention auto-encoder, to obtain cell embedding and gene embedding, SCEA can simultaneously achieve cell low-dimensional representation and clustering by performing various examinations to obtain the optimal value for each parameter, the presented result is in its most optimal form. To evaluate the performance of SCEA, we performed it on several real scRNA-seq datasets for clustering and visualization analysis. The experimental results show that SCEA generally outperforms several popular single-cell analysis methods. As a result of using all available data- sets, SCEA, on average, improves clustering accuracy by 4.4 percent in ARI Parameters over the well-known method scGAC. Also, the accuracy improvement of 11.65 percent is achieved by SCEA, compared to the Seurat model.
SCEA Workflow, The model consists of a basic MLP neural network with multiple layers and a graph attention neural network used for final dimensionality reduction. The reduced dimensionality of the graph will be used for clustering with the KMeans algorithm. KL loss represents the Kullback Leibler divergence and MAE is the Mean Absolute Error
DBSC is a Persian domain-based spell correction model that I developed, which was trained on data crawled from the Persian Wikipedia page. The model is a combination of a modified version of the Parsbert masked language model and edits distance to correct misspelled words. The DBSC model was evaluated by automatically creating noisy data and assessing how well our model and the fasttext baseline model could correct the created misspelled words. For more information about the project, please see the Github page of DBSC project.
In this project, we worked on medical images of the brain. We used the pre-trained vgg16 network and trained it on medical images. The transfer learning approach provided significant results. Moreover, using the Class Activation Map module, we discovered that our model colored only the parts which specify the category of the photo, proving the effectiveness of the transfer learning approach. Check out the project.
- Saeedeh Akbari Rokn Abadia, Seyed Pouria Laghaee, Somayyeh Koohi."An optimized graph-based structure for single-cell RNA-seq cell-type classification based on non-linear dimension reduction" published at BMC Genomics. DOI: https://doi.org/10.1186/s12864-023-09344-y