Skip to content

Latest commit

 

History

History
744 lines (553 loc) · 33.7 KB

Summary.md

File metadata and controls

744 lines (553 loc) · 33.7 KB

Contents Summary

Molecular Biology Concepts

Evolution has three main components:

  • Inheritance
    • Passage of characteristics from parents to offspring.
    • Determines most of the structure and functions of the organism.
    • The amount of variation passed from one generation to the next one is very small.
  • Variation
    • Occurs with mutations, sexual recombination, random changes of genetic material.
  • Selection
    • Reflects the fitness of the organism to adapt to the medium.
    • This fitness capacity is expressed through reproduction, i.e. parents transmit their fitness to their offspring.

Phylogenetics is the division of biology that studies evolutionary divergence and relationship between organisms, based on two important important concepts:

  • Similarity: Measures the resemblance and differences between organisms without taking into account any contextualization.
  • Homology: Investigates the common ground between the organisms and if they share any ancestral characteristics. Find the point in the evolutionary tree that they started to diverge.

The Micro perspective: The cell

The cell is the unit of life. Each cell derives from another cell and contains all the necessary information to replicate itself. All cells have some common features, such as its composition (70% water, 23 % macro-molecules and 7% small molecules).

Organisms are categorized according to their cell type:

  • Prokaryotes
    • No nucleus or internal membranes.
    • Lacks other membrane-bound organelles.
  • Eukaryotes
    • Nucleus (membrane-enclosed DNA).
    • Internal membranes.
    • Organelles inside the cell that play different and specific roles.

And according to the number of cells:

  • Unicellular
    • Prokaryotes: Bacteria, Archaea.
    • Eukaryotes: baker yeast.
  • Multicellular
    • Eukaryotes: animals, plants, fungi

What defines a cell?

  • Proteins perform most of the functions in the cell, they have catalytic and structural functions, from sensors and signaling to promoting chemical reactions (catalysis).
  • Enzymes are proteins that convert cellular molecules in other types of molecules necessary for the functions of the cell, like generating energy.

DNA and RNA are composed of nucleic acids. Proteins are composed of amino- acids.

Other important organic molecules:

  • Carbohydrates, store energy (simple - immediate energy demands, complex - long term storage of energy).
  • Lipids, make part of the plasma membrane and also store energy and are involved in signaling.

Other components of the cell:

  • Mitochondria and Chloroplasts are cellular organelles involved in the production of energy.
  • Ribosomes are large and complex molecules composed by a mixture of proteins and genetic material. Their function is to assemble proteins.

Cells form tissues that themselves form organs, and eventually entire organisms.

Information transfer in the cell: nucleic acids

Both DNA and RNA are polymers composed of four nucleic acid units, called nucleotides or bases.

  • Adenyne (A) and Guanine (G), belong to one group (purynes).
  • Cytosine (C) and Timine (T) and Uracil (U), belong to another group (pyrimidine).

Timine only exists in DNA and Uracil is only found in RNA, the other three bases exist in both.

The DNA is composed of two complementary strands, in a double-helix structure, due to connections established between the bases in both strands.

  • Adenine and Timine (A == T), connected by two hydrogen connections
  • Guanine and Cytosine (G === C), connected by three hydrogen connections

Chains are antiparallel because they are connected in opposite directions.

Genetic material

  • Genome: an organism’s genetic material (complete set of DNA).
    • human genome has 24 distinct chromosomes.
    • Each chromosome contains many genes.
    • In Prokaryotics existis in the form of a circular chromosome located in the cytoplasm.
    • In Eukaryotes is found in the nucleus and is tightly packaged into linear chromosomes.
  • Gene: a discrete units of hereditary information located on the chromosomes and consisting of DNA and encode instructions on how to make proteins.
  • Genotype: The genetic makeup of an organism.
  • Phenotype: the physical expressed traits of an organism.

Genetic material

Protein Synthesis

Cellular DNA contains instructions for building the various proteins the cell needs to survive:

  1. To manufacture these proteins, specific genes within its DNA must first be transcribed into molecules of mRNA (Transcription);
  2. these transcripts must be translated into chains of amino acids (Translation),
  3. fold into fully functional proteins.

In most eukaryotic genes, coding regions (exons) are interrupted by noncoding regions (introns). During transcription, the entire gene is copied into a pre-mRNA, which includes exons and introns. During the process of RNA splicing, introns are removed and exons joined to form a contiguous coding sequence. This "mature" mRNA is ready for translation.

All cells in a multicellular organism contain the same set of genetic information (Genome).

The differences in the abundance of the RNA (Transcriptome) determines the cell specificity.

Alternative Splicing

Alternative splicing (AS), the process in which the exons of the pre-RNAs are spliced in different combinations to produce distinct mRNA that lead to structurally and functionally protein variants.

Combinatorial splicing leads to the generation of of multiple isoforms from a single gene.


Basic Processing of Biological Sequences

Open Reading Frames

The translation of a protein sequence occurs for the coding region of the gene. This region start with a start codon (ATG) and stops when one of the stop codons is found.

A reading frame is a way of dividing a DNA (or RNA) sequence into a set of consecutive non-overlapping triplets or codons. Recall that a sequence may have 6 reading frames: 3 in one strand: +1, +2, +3 starting at position 1, 2 and 3 respectively of the sequence (in python strings at index 0, 1 and 2).

An open reading frame is a reading frame with the potential to be translated into protein (sarting with start codon - M, Meteonin - and the stop codon - A).

Putative proteins are the result of analysing all open reading frames of a given DNA sequence.


Finding patterns in Sequences

Examples where finding patterns might be useful:

  • mRNA contain in their promoter region special signal for the binding of the transcription factors that regulate gene expression.
  • Proteins contain sub-sequences, also called domains, that determine the function of the protein, including the binding sites for ligands or structural domains that determine the structure of the protein.

Sequence motifs are relatively short sub-sequences, shared among several (related) sequences that are presumed to have a similar biological function.

Hamming distance between two sequences, with the same size, are the number of different characters between the 2 sequences.

Example, where Hamming distance is 3.

Q:ATTACGAT
    | | |
T:ATCAGGTT

Improved search rules, that the Booyer-Moore algorithm includes:

  • Bad-Character Rule: The search can advance the pattern to the next occurrence of the symbol in the sequence at the position of the mismatch.
  • Good Suffix Rule: In case of a mismatch we can move forward to the next instance in the pattern of the part (suffix) that matched before of the mismatch.

Prosite is a database of protein families and domains. They analyze sets of related sequences to identify regions of sequence similarity. Prosite patterns are described using a syntax similar to the RE syntax:

# Convert a Prosite Pattern to a Regex Pattern
def prosite_to_regex(prosite):
    converter = {
        '-': '',
        'x': '.',
        '(': '{',
        ')': '}',
        '{': '[^',
        '}': ']'
    }

Restriction enzymes are proteins that cut the DNA in regions that contain specific sub- sequences or motifs. They are a very useful tool in molecular biology as they allow to create restriction maps useful for cloning and sequencing techniques.

IUB code: way of representing ambiguous sequences, using other characters.

iub_to_re = {
    "A":"A",
    "C":"C",
    "G":"G",
    "T":"T",
    "R":"[GA]",
    "Y":"[CT]",
    "M":"[AC]",
    "K":"[GT]",
    "S":"[GC]",
    "W": "[AT]",
    "B":"[CGT]",
    "D":"[AGT]",
    "H":"[ACT]",
    "V":"[ACG]",
    "N":"[ACGT]"
}

Pairwise Sequence Alignment

Bioinformatics relies on the assumption that biological sequences with high similarity share similar functions.

Phylogenetics is the division of biology that studies evolutionary divergence and relationship between organisms, based on Similarity (resemblance and differences between organisms) and Homology (common ground between the organisms and if they share any ancestral characteristics).

The edit distance between 2 sequences is the minimum number of editing operations to transform one sequence into the other. Operations:

  • Substitution: Replace symbols - T -> C.
  • Deletion: Delete 1 symbol - C -> _.
  • Insertion: Insert 1 symbol - _ -> C.

Sequence alignment is a way of arranging the biological sequences to identify regions of similarity. Such regions may indicate a functional, structural, or evolutionary relationships between the sequences.

A global alignment aligns two sequences from beginning to end, aligning each letter in each sequence only once. An alignment is always produced. (Needleman-Wunsch algorithm)

A local alignment maximizes the alignment of the parts of the sequences that share similarity. Finds the best aligned subsequence. An alignment may not be produced if no sufficient similarity is found. (Smith-Waterman algorithm)

Substitution matrices give a score for each substitution of one amino-acid by another. BLOSUM (BLOcks SUbstitution Matrix) matrix is a substitution matrix used for sequence alignment of proteins.

In pairwise sequence alignment we try to arrange two sequences so that the number of matching characters is maximised.

Dynamic Programming for Global Alignment:

Fill the score matrix cell by cell, using the value of adjacent cells to reach the target cell.

S[i][j] = max(S[i-1][j-1] + sm(a[i], b[j]), S[i-1][j] + g, S[i][j-1] + g) for all 0 < i <= n and 0 < j <= m

Matrix is filled left to right and top to bottom. To calculate S[i][j] you need to have calculated: S[i-1][j], S[i][j-1], S[i-1][j-1].

  • Iteration example 0:
gap = -8
seq1 = 'PHSWG'
seq2 = 'HGWAG'
gap H G W A G
gap 0 -8 -16 -24 -32 -40
P -8
H -16
S -24
W -32
G -40

Additionally, keep a trace-back matrix (T) to keep all the possible optimal moves at each cell. From T, one can recover the optimal alignment - start from lower-right cell and trace-back to upper-left cell.

Trace-back example

Assuming one has the trace-back matrix:

DIAGONAL = 1
VERTICAL = 2
HORIZONTAL = 3
gap H G W A G
gap 0 3 3 3 3 3
P 2 1 1 3 1 3
H 2 1 1 1 1 1
S 2 2 1 1 1 3
W 2 2 2 1 3 3
G 2 2 1 2 1 1

We start at TB[G][G]. There we have a 1, so we must proceed in the diagonal to TB[W][A] so our sequences look like:

G
G

Remember that since we are starting from the bottom right and going to the upper left in a trace-back style, we must also fill our alignment sequences in the opposite natural order.

Next, we find a 3 so we must move in the horizontal and reach TB[W][W]. Since we only decrease our column without decreasing the row, we are only moving through the second sequence. Our first sequence of the alignment will get a gap.

-G
AG

Next we find 1 so we reach TB[S][G].

W-G
WAG

Next we have a 1 so we reach TB[H][H].

SW-G
GWAG

Next we find 1 so we reach TB[P][gap].

HW-G
HWAG

Next we find a 2, so we must move in the vertical and reach TB[gap][gap]. Since we only moved in the first sequence, the second sequence will get a gap.

PHW-G
-HWAG

We have reached TB[0][0] and so we have finished!

Dynamic Programming for Local Alignment:

Find the best partial alignment of the sub-sequences from the 2 sequences.

S[i][j] = max(S[i-1][j-1] + sm(a[i], b[j]), S[i-1][j] + g, S[i][j-1] + g, 0) for all 0 < i <= n and 0 < j <= m

For the optimal alignment, one now starts in the cells with highest score.

Now, the trace-back matrix T contains 4 possible values: three previous values and an extra value to the cases where the alignment is terminated (correspond to cells in S with 0).

Trace-back example

Assuming we have the score matrix:

gap H G W A G
gap 0 0 0 0 0 0
P 0 0 0 0 0 0
H 0 8 0 0 0 0
S 0 0 8 0 1 0
W 0 0 0 19 11 3
G 0 0 6 11 19 17

And we have the correspondent trace-back matrix:

gap H G W A G
gap 0 0 0 0 0 0
P 0 0 0 0 0 0
H 0 1 0 0 0 0
S 0 0 1 0 1 0
W 0 0 0 1 3 3
G 0 0 1 2 1 1

We start at the position of the cell with maximum score. In this case there are 2 cells with maximum score, so we will end up with 2 possible best alignments.

  • For alignment_one, we start at TB[G][A] nd we have a 1 so we move diagonally to TB[W][W]. There we have a 1 as well so we move to TB[S][G]. Once again a 1, to TB[H][H]. Again a 1 to TB[P][gap] were we have a 0, so we stop. We end up with the following alignment evolution:
G
A
WG
WA
SWG
GWA
HSWG
HGWA
  • For alignment_two we start at TB[W][W] and then, form there, its actually the exact same alignment as alignment_one, so:
HSW
HGW

Searching for Similar Sequences in Databases - BLAST

In Bioinformatics to infer the function of an unknown sequence, one has to scan large databases of sequences and compare the query sequence against all the sequences in the database, selecting those with higher degree of similarity.

Pairwise sequence alignment algorithms are not efficient enough (quadratic complexity & memory consuming because of 2 matrixes) when it is necessary to scan very large sets of sequences, so BLAST - Basic Local Alignment Search Tool - is used.

BLAST can be used to infer functional and evolutionary relationships between sequences as well as help to identify members of gene families.

BLAST Concepts:

  • Query sequence: sequence that will be processed with BLAST (to know more about).
  • Target sequence: sequence in the database that was matched with the query sequence.
  • Database (D): database of sequences where the search is done.

BLAST is an heuristic approach based on the idea of K-indexing.

K-mer Indexing concepts:

  • Database Pre-processing: Scan every word of length K and keep it in a hashmap.
  • Query sequence scan: scan every k-word in Query and get its location in the hashmap.

BLAST Algorithm:

  1. Seeding: find common subwords between the query sequence and the database sequences (using the K-mer Indexing) -> seeds.
    • Window scanning of Q to generate K-words: L1-set
    • Find neighborhood words for each k-word until threshold T: L2-set
    • Merge L = L1 U L2
    • Look into H table where L words occur: seeds
  2. Extension: starting from seeds, extend alignment in both directions -> high-scoring segment paris (HSP).
    • BLAST 2.0. allows alignments with gaps. If two non-overlapping hits are found within distance A of one another on the same diagonal, then merge the hits into an alignment and extend the alignment in both directions until the running alignments score has dropped more than X below the maximum score yet attained.
  3. Evaluation: assess the statistical significance of each HSP.
    • The Expect value (E) is a parameter that describes the number of hits one can "expect" to see by chance when searching a database of a particular size. It decrease exponentially as the Score (S) of the match increases. The lower the E-value, or the closer it is to zero, the more "significant" the match is.

Blast Explanation

BLAST can be summarized in:

  1. Remove regions of low complexity (e.g. sequence repeats) from query sequence.
  2. Obtain all possible “words” of size w (a parameter of the algorithm), i.e. sub- sequences of length w occurring in the query sequence;
  3. For each word from the previous step, compile the list of all possible words of size w that can be defined in the allowable alphabet, whose alignment score (with no gaps) is higher than a threshold T (parameter of the algorithm);
  4. Search in all sequences from the database, all occurrences of the words collected in the last step, which represent matches (hits or seeds) of size w between the query and one of the database sequences;
  5. Extend all hits from the last step, in both directions, while the score follows a given criterion (typically, the criterion is dependent on the size of the extension);
  6. Select the alignments in the previous step with highest scores, normalized for its size (these are named the high-scoring pairs - HSPs).

Consideration

Notice however, that the teacher MyBlast implementation only creates a hash table out of the query sequence - build_map() function. Regarding the database, he iterates over it, calling the get_hits() method for each stored sequence.

Exercises

Original get_hits() function:

def get_hits (seq, m, w):
    res = [] # list of tuples

    for i in range(len(seq)-w+1):
        subseq = seq[i:i+w]

        if subseq in m:
            l = m[subseq]

            for ind in l:
                res.append((ind, i))

    return res

get_hits() allowing at most 1 mismatch, meaning returns it all hits that have w or w-1 matches:

def hamming_distance(seq1, seq2):
   assert len(seq1) == len(seq2), "Sequences can not have different lengths"

   return sum(
       [1 for i in range(len(seq1)) if seq1[i] != seq2[i]]
   )

def get_hits (seq, m, w):
   res = [] # list of tuples

   for i in range(len(seq)-w+1):
       subseq = seq[i:i+w]

       for word in m.keys():
           if haming_distance(subseq, word) <= 1:

           for ind in m[word]:
               if (ind, i) not in res: # No duplicates
                   res.append((ind, i))

   return res

Multiple Sequence Alignment (MSA)

Generalization of the Alignment Problem to multiple sequences (pairwise alignment problem (PSA) for N > 2 sequences).

Possible utilities of MSA:

  • identify recurrent motifs across a family of protein sequences
  • study homology relations between an evolving gene (Phylogenetic analysis)
  • assess conservation on secondary and tertiary protein structures
  • epidemiological studies to understand the mutation rate of a gives species strain.

MSA can be used to start phylogenetic analysis. The phylogenetic tree describes the distance between the sequences under analysis. From the alignment, each column shows if there are conservation of the residues (amino-acids), mutations or divergence from the common ancestor.

Compared to the PSA, we now have multiple symbols per column. The sum of pairs (SP) method considers all possible pairs of symbols within the column and sums them. Eventually, it counts the number of gaps only once.

With many sequences, dynamic programming becomes inviable and therefore Heuristic approaches are used. They represent a trade-off between speed and optimality. Some Heuristic approaches are:

  • Progressive: start by aligning the two most similar sequences and iteratively add the other sequences to the alignment.
  • Iterative: consider an initial alignment and then improve it by adding, removing or moving gaps.
  • Hybrid: combine both strategies and use complementary information (e.g. protein structural information, other good alignments)

The CLUSTAL algorithm is a classic MSA method that is at the basis of many MSA algorithms. CLUSTAL approach (Progressive approach):

  1. Calculate pairwise alignment of all sequences and build a similarity matrix
  2. Select the most similar sequences to form the basis of the MSA; the order of the sequences follows that of the guide tree;
  3. To add more than two sequences we need to create, from the existing alignment, a summary of the content on each column.
    • Consensus representation: each column is represented by the most common character;
    • Frequency representation: each column is represented by the frequency of the characters, also called profile.
  4. At each iteration align the new sequence with the profile of the current MSA.
    • The column score is the weighted average of the scores of all possible pairs.
    • Once a gap is added to the alignment it will prevail as gap throughout the next alignment steps.
    • In some situations, we may need to join to sub-alignments corresponding to two different sub-branches of the tree. Requires joining two profiles in a generalisation of the above process.

CLUSTAL Overview

Cons of Heuristic approaches:

  • If wrong decisions are made early in the alignment they will be propagated and not corrected afterwards.
  • Worst results occur when sequences have low similarity.

Phylogenetic Analysis

Phylogenetics studies the evolutionary history and relationship among individuals or species while Phylogenetics trees illustrate the relationship between these individuals.

Phylogenetics trees

  • Leaves are sequences (typically from different species or taxonomic categories)
  • Internal nodes represent common ancestors of the sequences.
  • The structure of a rooted tree may be represented by clusters.
  • The height of the nodes in the tree represent a measure of time (moving from the root to the leaves).
  • The length of the branch represents the evolutionary distance between the ancestor and the species at the node. This is captured by the number of sequence changes between one level and the next level of the tree.
  • Trees without a root illustrate the relation between the leaves without explicitly inferring the common ancestor.

The Phylogeny Problem

Assume we are given a set of sequences evolutionarily related, i.e. with a common ancestor. The problem: infer the best possible evolutionary tree. This is an optimisation problem, since the number of possible trees increases exponentially with the number of input sequences, and therefore requires an objective function. There are three main algorithms:

  • Distance-based algorithm: compute a distance matrix based on the pairwise distances of the sequences. Derive trees consistent with the distances from the matrix.
  • Maximum parsimony: search for trees that try to minimize the number of mutations (in internal nodes) to explain the variability of the sequences. Based on MSA of the input sequences. Use certain columns in the alignment that are informative of the possible phylogeny.
  • Statistical/Bayesian: probabilistic models for the occurrence of different types of mutations in the sequences. Score trees based on their probability searching the most likely trees that explain the sequences according to the assumed model.

Distance-based methods

Rely on measuring the consistency of the distances between the leaves in the tree (sequences) and the distances derived from sequence similarity (alignment). The structure of the tree and the length of the branches connecting the nodes reflect the pairwise distances between sequences.

Distance is the reverse of similarity (e.g. percentage of columns in the alignment with mismatches or gaps)

Score function for Distance-based methods

  • S: set of input sequences
  • T: tree
  • dij(T): distances of the leaves representing sequences i and j in the the tree.
  • Dij: distance between sequences I and j in the input matrix D given from sequence * alignment.

Ultrametric tree: the distance between all leaves and the root is the same. The height of each leaf is 0. Thus, duw = h(w) -> duv = 2 * h(w).

As the number of sequences increases the solution space also increases exponentially. Heuristic methods need to be applied to obtain solutions in reasonable time. Hence, use Unweighted Pair Group Method Using Arithmetic Averages - UPGMA, which is based on agglomerative hierarchical clustering algorithms.

Clustering Algorithm (uses Ultrametric Tree):

  • Consider each sequence (tree leaf) as its own cluster. height = 0 in the tree.
  • Merge the pair of closest sequence/clusters (minimum value in the matrix D); join these sequences creating an internal node. The height = half of distance between sequences. These sequences form a cluster.
  • Distance of a cluster to the remaining sequences is the average of the distances. Update distance matrix D: remove cols and rows of the connected sequences. Add row and col for the new cluster.
  • Iteratively: find pairs of clusters with minimum distance and repeat: join clusters, add internal node to the tree with the given height and update D.
  • Stop when all sequences are within a single cluster that corresponds to the root of the tree.

Clustering Algorithm

Practical example of UPGMA:

Having the sequences:

s1 = A-CATATC-AT-
s2 = A-GATATT-AG-
s3 = AACAGATC-T--
s4 = G-CAT--CGATT

Distance Matrix (using Hamming Distances):

  • Iteration 0
s1 s2 s3 s4
s1 0
s2 3 0
s3 4 6 0
s4 5 8 9 0

The most similar sequences are the s1 & s2, and so we start our tree, with the height being 3/2:

     ___
1.5 |   | 1.5
    |   |
   s1   s2

Now, lets compute the new distance matrix, remembering that:

Imgur

|A| represents the cluster size of A, e.g. |s1, s2| = 2 and |s1, s2, s3, s4| = 4.

  • Iteration 1

  • d(s1 U s2), s3 = (1 * 4 + 1 * 6) / (1 + 1) = 5

  • d(s1 U s2), s4 = (1 * 5 + 1 * 8) / (1 + 1) = 6.5

s1, s2 s3 s4
s1, s2 0
s3 5 0
s4 6.5 9 0

The most similar sequences are the s1,s2 & s3, and so we update our tree, with the height being 5/2:

        _____________
  2.5  |             |
     __|__           | 2.5
1.5 |     | 1.5      |
    |     |          |
   s1     s2         s3
  • Iteration 2

  • d((s1, s2) U s3), s4 = (2 * 6,5 + 1 * 9) / (2 + 1) = 7.33

s1, s2, s3 s4
s1, s2, s3 0
s3 7.33 0

The only available sequences are the s1,s2 & s3, and so we update our tree, with the height being 7.33/2:

                         |
               __________|________
         3.66 |                   |
              |                   |
        ______|______             |
  2.5  |             |            | 3.66
     __|__           | 2.5        |
1.5 |     | 1.5      |            |
    |     |          |            |
   s1     s2         s3           s4

Graphs and Biological Networks

The elements of the cell form networks that are significantly different from random networks. Examples of cell networks include:

  • Protein-protein interactions;
  • Metabolic;
  • Signalling and Metabolic networks;
  • Protein phosporylation*;
  • Genetic interactions;
  • Co-expression networks;
  • Protein-DNA interactions;

The cell is formed by the interplay of networks and emerge as the sum of the interaction of its different elements. It is a network of networks.

Graphs can be:

  • Undirected if edges are unordered
  • Directed or digraph if edges have an orientation, i.e. pairs are ordered
  • Weighted if numerical weights are associated to edges.

Matrices or adjacency lists are the typical way to represent graphs.

Matrices Representation

All possible combination of vertices are represented.

  • Undirect graph: the rows and cols represent the nodes. Cell(i,j) represents an edge between node i and j: 1 if connected and 0 otherwise.
  • Direct graph: rows represent the origin node and cols the destination node. Cell(i,j) represents an edge between node i and j: 1 if connected and 0 otherwise.
  • Weighted graph: Cell(i,j) represents the weight of the edge between node i and j. 0 if not connected.

Adjacency lists Representation

Represent only the existing edges. Each vertex is linked to a list with associated neighbor nodes.

  • Direct graph: list for vertex v includes all destination nodes for the edges where v is the origin.
  • Undirect graph: the edge only exist for one of the directions.

Some Network concepts:

  • Degree: how many links a node has.
  • Degree In: ho many links a node has entering it.
  • Degree Out: ho many links a node has leaving it.
  • Shortest path: path with the smallest number of links between selected nodes. Distances are measured as the length of the path.
  • Degree distribution P(k): gives the probability that a selected node has exactly k links. P(k) is obtained by counting how many nodes have k = 1,2,... links and dividing by the total number of nodes N. The degree distribution will allow to discriminate between different types of networks.
  • Clustering coefficient: indicates how connected are the the nodes on the network. If a node A is connected to B and B is connected to C, then it is highly probable that A is connected to C. The clustering coefficient of a node, gives the number of triangles that go through that node

The last two measures capture the features and characteristics of the network and therefore allow to compare and classify the various networks.

Different type of networks according to degree distribution:

  • Scale free network: the degree distribution approximates a power-law distribution where P(k) ~ k-a
    • Most networks on the cell are scale-free.
  • Random network: P(k) decreases exponentially, which indicates that nodes that deviate from average are extremely rare.
  • Hierarchical network:
    • The starting point is a small number of four nodes highly linked.
    • Communication between highly clustered models is maintained by hubs.
    • Scale free topology + modular structure
    • Clustering coefficient follows C(k) ~ K-1 is the most important signature for this type of networks.

Two strategies to traverse the graph:

  • Breadth-First Search (BFS): starts with the source node, then visits all its successors, followed by their successors until all nodes are visited.
  • Depth-First Search (DFS): starts with the source node, explores its first successor, then its first successor until no further exploration is possible. Backtracks to explore further alternatives.

Not all nodes are equally significant. Identification of nodes that are most important is a central task in network analysis.

Identification of hub nodes provides information on important properties of the network, like the robustness and resistance to failure. In biological networks a hub node in a gene-gene or gene-protein interaction network may reveal an important regulator, for instance a transcription factor that regulates many other genes.


High-throughput sequencing applications

Sequencing Technologies

Presents in slides the typical applications of massive (also called next-generation or high-throughput) sequencing technologies.

DNA Sequencing: process of determining the order of nucleotides in DNA.

Exercises

Exercise #1: RPKM (Reads per KiloBase per Million):

Used for single end RNA-sequencing.

Notice that:

  • Sequencing runs with more depth will have more reads mapping to each gene.
  • Longer genes will have more reads mapping to them.

RPKM = 10^9 * #reads / ((Σ #reads) * Len)

gene Len Reads RPKM
A 2 20
B 3 10
C 5 50
D 10 100

We can infer that gene A's sequencing runs have more depth than gene B, since they have similar gene size but A #reads is 2x bigger.

Σ #reads = 20 + 10 + 50 + 100 = 180

Example A <=> RPKM = 20 * 10^9/ (180 * 2) = 5 * 10^7

gene Len Reads RPKM
A 2 20 5.5
B 3 10 1.8
C 5 50 5.5
D 10 100 5.5

Exercise #3: Candidate mutations in sequencing read pile-up

CGACGACGACGACGAATGATGTATTATCGAGCGAGCGGCAGATGCTA
-----------------------------------------------
CGACGACGACGACGAATGAT    TATCGAGCGCGCGGCAGATG
 GACGACGACGACGAATGATG      CGAGCGCGCGGCAGATGCTA
 GACGACGACGACGAACGATG      CGAGCGCGCGGCAGATGCTA
         CGACGAACGATGTATTATCG
            CGAACGATGTATTATCGAGC
                   TGTATTATCGAGCGCGCGGC

Reads that do not match the reference genome:

CGACGACGACGACGAATGATGTATTATCGAGCGAGCGGCAGATGCTA
-----------------------------------------------
CGACGACGACGACGAATGAT    TATCGAGCGCGCGGCAGATG
 GACGACGACGACGAATGATG      CGAGCGCGCGGCAGATGCTA
 GACGACGACGACGAACGATG      CGAGCGCGCGGCAGATGCTA
         CGACGAACGATGTATTATCG
            CGAACGATGTATTATCGAGC
                   TGTATTATCGAGCGCGCGGC
                ^                ^