Programming for Biologist
Part #1: SAMParser.py
Next-generation sequencing technologies are making a large impact in Biology today. The
ability to sequence large genomes and RNA molecules from cells in different conditions has
allowed biologists to propose many different experiments. At the root of all the analysis the
sequences have to be first aligned to the genome so you know where the sequence comes
from.
For this assignment you will parse the results of an alignment ( called a SAM/BAM file ) and
compare it to a file that contains gene annotations (GTF) .
The documentation for the SAM/BAM file can be found here. (https://samtools.github.io/htsspecs/SAMv1.pdf). The name of the BAM file we will be using is sample_sort.bam. The two
columns of interest for our purpose are column 3, the name of the chromosome, and column 5,
the start position of the alignment. Fortunately for us there is a library called PySAM
(https://pysam.readthedocs.io/en/latest/) so we can parse the file easily. In order for us to use all
the functionality of PySAM library, we need to also use an index file sample_sort.bam.bai. An
index file is just like the index in a textbook, it tells the program where in the file the different
chromosome’s alignments start.
Now the GTF file (Arabidopsis.gtf) is a tab-separated file already sorted. It contains the
chromosome in the first column and the start and end positions of an exon in the 4th and 5th
column, and the gene name in the last column.
Remember that for certain genes there are more than one transcripts. Which means exons are
repeated. To resolve this we will use a simple rule, let’s just take the first transcript. So use only
the exons that belong to the first transcript, their ID ends with a .1
To determine the number of reads that match a gene, count the number of start positions from
the SAM file that fall between the start and stop coordinates of an exon that belongs to a gene.
You can use the count function or the fetch function in Pysam
(https://pysam.readthedocs.io/en/latest/api.html#pysam.AlignmentFile.count) to help you with
this.
Note - some reads can match multiple exons. To be technically correct, make sure you don’t
count the same read twice. But for this assignment we will not deduct points for not keep
tracking of reads that match more than one exon.
The output should be a matrix with two columns, one for the gene name and the other for the
number of reads that matched. Make sure to include all genes, even if they do not have a
match.
Part #2: Write your own KNN from scratch
You are given a matrix where 20 genes are used as biomarkers to identify what type of cancer
the patient has. This labeled data is training.txt.
You are also given a smaller set of data (test.txt) for the same 20 genes, but you will have to
predict the label ( the type of cancer).
For each individual in the test set, use euclidean distance to find the closest k members of the
individual in the Training set. Predict the label based on which label is in the majority among the
labels of its neighbors. The output is the test set with the predicted labels.
KNN.py will accept four arguments:
1. - r The training data
2. - t The test data
3. - k The k to be used
4. - o The name of the output where the predictions should be saved.