Kenneth Y. Wertheim's Projects
Working on the Galaxy platform, I aligned three sets of targeted re-sequencing data to a reference genome called hg19, called the variants, and annotated them.
The Python program main.py aligns DNA sequencing reads with an excerpt of human chromosome 1 by dynamic programming, finding the minimum edit distance in each case. It also finds the overlaps between sequencing reads from Phi-X.
The Bash script is a pipeline for variant calling.
The Bash script 'analysis.sh' uses samtools and bedtools to quantify the relations between a set of sequencing reads and their alignments on the genome of one strain (wu_0_A) of the plant Arabidopsis thaliana.
The Bash script 'analysis.sh' analyses the files in an archive and reports key statistics about the genome of Malus domestica.
The Bash script 'analysis.sh' runs the RNA-seq Tuxedo pipeline: TopHat, Cufflinks, Cuffcompare, Cuffmerge, and Cuffdiff.
The R script uses the AnnotationHub package to obtain data on human CpG islands and histone modifications (H3K4me3 and H3K27me3); and the GenomicRanges package to extract basic statistics about the epigenome from the data.
The R script uses Bioconductor tools to study various epigenomic features and the promoters of chromosome 22 in hg19.
The R script uses Bioconductor tools to obtain and analyse microarray (gene expression and methylation) and RNA-Seq data.
The R script uses Bioconductor tools to obtain and analyse RNA-Seq, gene expression microarray, and epigenomic data.
The Python program main.py aligns DNA sequencing reads with an excerpt of human chromosome 1 using the naive exact matching and Boyer-Moore algorithms, as well as k-mer indices (substrings and subsequences) and the pigeonhole principle.
The files in this repository constitute a multi-approach and multi-scale model of CD4+ T lymphocytes. Running the main.m file within the MATLAB environment will start a stochastic simulation of CD4+ T cells responding to the input signals defined in antigen.mat.
The Jupyter notebook presents a DCGAN built for the purpose of generating new images of human faces, as well as the procedures for training it with a processed CelebA dataset.
The Jupyter notebooks present the procedures I followed to build a facial keypoint detection system.
The Python program main.py contains the shortest common superstring algorithm, the greedy shortest common superstring algorithm, and an accelerated version of the greedy shortest common superstring algorithm. They can be used to assemble a genome from its sequencing reads.
This program extracts insights from multi-FASTA files.
The Python program checks the quality of DNA sequencing reads and aligns reads with a genome using the naive exact matching algorithm and modified versions of it.
This repository contains the first multicellular model of neuroblastoma, built for the PRIMAGE project.
These notebooks present a series of attention models for natural language processing, including an encoder-decoder model (translator) comprising LSTMs and the scaled dot product attention mechanism, a transformer decoder (text summariser), a transformer encoder (question answering) based on the BERT model, and a reformer (chatbot).
These notebooks introduce the fundamentals of natural language processing, including text preprocessing techniques, logistic regression, Naive Bayes classification, word vectors (embeddings), principal component analysis, the bag-of-words model, the k-nearest neighbours algorithm, and locality-sensitive hashing.
These notebooks present a hidden Markov model (part-of-speech tagging), an n-grams language model (auto-complete system), and a CBOW model (computing word embeddings), as well as supporting techniques such as dynamic programming (autocorrect system), the Viterbi algorithm, and the perplexity score.
These notebooks present a series of deep neural networks for natural language processing, including a feedforward network, an RNN with GRUs, an LSTM, and a Siamese network. Their applications are sentiment analysis, next character prediction, named entity recognition, and detecting duplicate questions respectively.
This repository contains two sets of code. One is used to simulate neuroblastoma's clonal evolution in the presence of vincristine and cyclophosphamide. The other is used to find the optimal chemotherapy schedule given an initial clonal composition.
This algorithm builds a series of paediatric cancer classifiers, including a decision tree, a naive Bayes classifier, support vector machines, an ensemble method (Adaboost), and a multilayer perceptron. It also implements hierarchical clustering and principal component analysis.
This repository contains various C++ programs, including recursion, dynamic programming (memorisation and tabulation), and genetic algorithms (binary and continuous).
These algorithms assess the potential of a multi-component system to form Turing patterns. The reaction-diffusion model describing the system is analytically intractable. The solution employed is to explore the parametric space one point at a time.